| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Division of Molecular Genetics, Institute of Biomedical and Life Sciences, University of Glasgow, Glasgow, United Kingdom
ABSTRACT I. WHAT IS INTEGRATIVE PHYSIOLOGY? II. WHICH TRANSGENIC MODEL? A. Choosing Your Model Organism Is a Trade-off B. Real Physiology in Model Organisms 1. Shaker 2. Learning and memory 3. Circadian rhythms 4. The inositol trisphosphate receptor III. WHAT IS FUNCTIONAL GENOMICS? IV. WHY GENOME PROJECTS NEED MODEL SYSTEMS V. WHAT IS REVERSE GENETICS? A. Reverse Genetics Demands a Genetic Model B. Even Human Genomics Requires Simple Models C. Other Routes to Functional Analysis VI. THE ''PHENOTYPE GAP'' VII. THE DROSOPHILA MELANOGASTER MALPIGHIAN TUBULE A. History of Malpighian Tubule Physiology B. How Is the Drosophila Malpighian Tubule Organized? 1. The classical view 2. Enhancer trapping as a tool to understanding organization 3. The principle of enhancer trapping 4. Enhancer trapping can detect tubule-specific genes 5. Enhancer trapping reveals regional specialization 6. Enhancer trapping reveals multiple cell types 7. Genetic domains can be quantified by counting nuclei 8. The GAL4/UAS system 9. Correlating functional and genetic maps 10. Phylogenetic scope of enhancer trapping technology 11. Emulating enhancer trapping in other organisms VIII. PHYSIOLOGY AND MOLECULAR GENETICS OF ION TRANSPORT IN RENAL FUNCTION: V-ATPASES A. Drosophila vha55 Is Single Copy and Located at 87C B. The 87C Region Has Been Subjected to Intense Genetic Analysis C. P-element LacZ Reporter Reveals V-ATPase Expression Patterns D. Correlating Functional and Genetic Maps: V-ATPase Expression Levels E. V-ATPases Are Often Found in Specialized Cell Types F. The SzA Locus Provides Multiple Alleles of vha55 1. An epithelial phenotype in larvae homozygous for lethal mutations in V-ATPase G. A Human Renal and Auditory Phenotype Is Associated With a Plasma-Membrane V-ATPase IX. HOW DO CHLORIDE AND WATER CROSS THE TUBULES? A. Leucokinins Selectively Increase Chloride Conductance B. Patch Clamp Reveals Maxi-Chloride Channels in Tubules C. Self-Referencing Electrode Analysis Shows That Chloride Moves Through Stellate Cells D. Tubules Contain CLC Chloride Channels E. Bartter Syndrome Type III F. How Does Water Cross the Tubule? G. Cation and Anion Transport Are Spatially Separated X. PHARMACOLOGY AND CELL SIGNALING MECHANISMS A. Cell-Specific Calcium Signaling B. Neuropeptide-Stimulated Calcium Signaling C. Calcium Channels D. NO/cGMP Signaling E. Cross-talk Between NO/cGMP and Calcium Signaling F. New Targets for Drug Discovery G. Insect Tubules as Targets for Selective Insecticides XI. IN SILICO MAPPING OF METABOLIC PATHWAYS A. Human, Mouse, and Gout B. Drosophila, Mice, Humans, and Gout C. Disorders of Purine Metabolism D. Does Drosophila Have a Xanthine Oxidase Mutant? E. Virtual Metabolomics: an Index of Possibilities F. Eye Color Mutants Have Interesting Causes XII. DROSOPHILA AS A MODEL FOR HUMAN DISEASE A. Can Drosophila Cure Human Disease? B. How Different Can You Be and Still Be the Same? C. Functional Genomics Offers a Future for Comparative Physiology 1. Vector biology 2. The lower tubule domain is conserved among Diptera D. Functional Genomics and Integrative Physiology Are Linked E. A Plan for Physiologists 1. How can genome projects help physiologists? 2. How can physiologists help genome projects? XIII. FUTURE PROSPECTS XIV. CONCLUSIONS
 |
ABSTRACT |
|---|
 Top NextReferences |
|---|
|
I. WHAT IS INTEGRATIVE PHYSIOLOGY? |
|---|
|
TopPreviousNextReferences |
|---|
|
These are enormously powerful techniques. However, there is a set of problems that these techniques cannot address, simply because they have taken the system under study away from the physiological context in which the question was first posed. For example, in the context of epithelial biology, how does an epithelium develop? How is the epithelium polarized so that different transport proteins are found on opposite sides? How are heterogeneous cell types specified in the tissue, and how do they interact? What signaling processes allow the tissue to coordinate its function?
The realization that analytical techniques, and in particular molecular biology, cannot give all the answers has led to a call for a new, "integrative" physiology that seeks to move from single molecule back to the whole organism (164). However, rather than throw the baby out with the bathwater, modern genetics provides tools that let us combine the strengths of the reductionist approach with the relevance of organism-level studies. Ideally, we would like to be able to manipulate specific genes in defined cells in particular tissues at specific life stages in the intact organism, as easily as cells can be transfected with DNA in vitro. Once the problem is phrased in this way, it is clear that a transgenic approach is needed.
|
II. WHICH TRANSGENIC MODEL? |
|---|
|
TopPreviousNextReferences |
|---|
There are fairly few organisms in which transgenesis is routine (Table 1). These tend to be termed "genetic model organisms," and in addition to transgenesis, there are frequently genome project resources associated. For biomedical research, it might seem that only the mouse is a realistic candidate for transgenics. It is a vertebrate and has a recognizable body plan, and reasonably well conserved physiology, compared with humans. However, its high biomedical relevance is offset by its very low genetic power. Although targeted mutagenesis by homologous recombination is possible, it is slow and very expensive, taking perhaps as much as 3 years. This makes it hard to both generate and analyze a transgenic line in a single grant cycle.
|
Although the mouse is, rightly, the biomedical model of choice, its genetic limitations are manifest. In contrast, Drosophila represents an ideal trade-off between human biomedical relevance and genetic power. It is no coincidence that Celera chose Drosophila as the warm-up genome to practice their techniques for humans. In fact, even simpler models have their place; just as nearly all we know about developmental genes was first established in Drosophila, so the bulk of our knowledge about the cell cycle was derived from yeast, and our understanding of DNA replication and gene regulation was pioneered by studies in the humble bacterium Escherichia coli. The relative merits of model organisms are nicely reviewed elsewhere (17). It is thus clear that good science is facilitated by a wise choice of model organism that is relevant to the question being posed, and a thrust of this article will be that the mouse is not necessarily the automatic choice for postgenomic physiology.
What is meant by genetic power? Targeted mutagenesis is certainly part of the equation, but there are other aspects, such as the facility with which forward genetics can be undertaken. Although mouse is the smallest and cheapest vertebrate to keep, the costs are far too high, and life cycle far too long, to imagine setting up genetic screens routinely. However, the pay-offs can be large. For example, an industrial/academic collaboration has produced a systematic ENU mutagenesis of mouse, with a view to uncovering genes involved in human neurological deficits (6, 16, 108, 117, 167, 213). An alternative approach is to draw on the many strains of laboratory rat that have defined vascular defects and to map (painstakingly) a variety of blood pressure, osmoregulatory, and renal phenotypes through huge back-crosses, to identify loci by a quantitative trait locus (QTL) approach (113, 230). It is highly likely that loci identified in hypertensive rats will be directly relevant to our understanding of hypertensive disease in humans. However, these high-profile exceptions tend to prove the rule that rodent forward genetics is so expensive and time-consuming that it can only exceptionally be undertaken. Should this lack of genetics be of concern to a physiologist? Perhaps surprisingly, there are whole swathes of physiological endeavor with roots in forward genetic screens of "simple" model organisms. The next section reviews some areas in which our understanding of basic physiology has been advanced by classical genetic studies in Drosophila melanogaster.
B. Real Physiology in Model Organisms
In the mid 1980s, the first voltage-gated sodium channels were cloned contemporaneously from rat (165), fly (200), and eel (166). There was massive excitement that all known ion channels matched the same structural template, with a six transmembrane motif repeated four times, presumably by two ancient gene duplication events (202). This argument was dramatically vindicated by the discovery of an "ancestral," six-transmembrane potassium channel in Drosophila (41). The route to this discovery was unconventional: whereas the original sodium channels had been cloned by molecular biochemical means (cDNA libraries were screened with degenerate probes derived from peptide microsequence of sodium channels purified by tetrodotoxin affinity), the Shaker channel was identified by forward genetics.
A serendipitous behavioral mutation, in which flies' legs shook under ether anesthesia (41), was analyzed physiologically and shown to be caused by an underactive A-type potassium current in muscle (201). The locus was identified by positional cloning (118) and shown to encode a quarter-sized channel, with six transmembrane domains. The identity of the locus was confirmed by showing that the gene was indeed mutated in Shaker mutants. Although the naïve interpretation would have been that flies, being primitive, carried an ancestral channel, this would have been far from the truth: Shaker homologs were rapidly identified in mammals (199). The beneficial interplay between physiology and genetics is nicely illustrated again; although Shaker encoded a potassium channel, A-type potassium currents remained in Drosophila Shaker mutants, suggesting that further related genes might exist (39). This led to a conventional library-screening search for related genes, identifying prototypes of three further Shaker subfamilies: Shal, Shab, and Shaw (39). These genes, in turn, were shown to have human counterparts (199).
In this case, there is no doubt that the Sh, Shab, Shal, and Shaw family of channels would have been discovered in time. However, the use of a genetic model organism advanced the field by at least a decade. Genetics and physiology can thus be potent fellows (202).
To most scientists, the conspicuous success of comprehensive forward screens in Drosophila has been the revolution in our understanding of developmental biology, in recognition of which Edward B. Lewis, Christiane Nusslein-Volhard, and Eric F. Wieschaus were awarded the 1995 Nobel Prize in Medicine. However, there are rare cases where elegant forward genetic screens have advanced our understanding of physiology. Here are two examples: learning and memory as well as circadian rhythms.
Although Drosophila is a relatively simple organism (104 neurons vs. 1012 in humans), it displays recognizable behaviors, such as learning and memory. With a foresight that rivaled the great developmental screens, a systematic olfactory learning screen was employed to demonstrate a role for cAMP in learning and memory.
The original screen, much copied since, identified a mutation in a locus (dunce) that was defective in a range of simple associative learning tasks, but essentially normal in other aspects of behavior (74). Parallel work on adenosine metabolism showed that dunce mutants had lower levels of cAMP phosphodiesterase (40, 45). Dunce was subsequently found to encode a cAMP phosphodiesterase that was expressed predominantly in those brain regions thought to encompass learning behavior (163). Support for cAMP as a major messenger in learning and memory was gained from the discovery that rutabaga, another learning mutant, showed aberrant adenylate cyclase activity (75), and the gene was subsequently shown to encode an adenylate cyclase (137).
In parallel with the Drosophila work, cAMP was shown to play a role in a simple habituation response, the gill withdrawal reflex, in the sea hare Aplysia (119). The Aplysia result both extended the phylogenetic scope of cAMP as a mediator in learning and memory and nicely illustrates the limitations of work in a nongenetic model, as it was hard to take the work forward rapidly in Aplysia. In Drosophila, it was possible to extend the results of the learning screen by crossing dunce to rutabaga and showing that in the double mutant, both learning behavior and overall cAMP levels were almost normal (78). Thus a defect in adenylate cyclase could be compensated for by a defect in the cognate phosphodiesterase. The principle of investigating genetic interactions has been extended in subsequent work. Another feature of a genetic model organism is the availability of transgenics, and this has also been brought to bear. Overexpression of a dominant negative pseudosubstrate inhibitor of cAMP-dependent protein kinase (cAK) has been shown to disrupt learning, implicating the kinase in the learning and memory pathways (71). In Drosophila, it is relatively easy to target expression of transgenes, and this has allowed further physiological analysis: by overexpression of wild-type rutabaga adenylate cyclase in different brain regions, it was possible to show that the critical regions for learning and memory were the ventral ganglion, antennal lobes, and median bundle (255). It is hard to imagine how such functional insights could be obtained by other means.
Another major screen for genes underlying basic behavior was that for mutants in circadian clock function. The archetypal clock gene, period, was first found in Drosophila (131). Although the structure of its encoded peptide was novel, it is the prototype of a large family since implicated in circadian function in mammals, and even humans (224).
The essence of the method is the design of the screen: only with a robust paradigm can a graduate student be expected to screen the 100,000 flies that might prove necessary to identify likely mutants. (The scale required also neatly illustrates the unsuitability of mouse for such screens.) In this case of Drosophila clock genes, the screen was suggested by a long history of classical work on molting and clock genes in insects; adult flies tend to emerge from the pupa (eclose) around subjective dawn (49). Therefore, if larvae are reared under a specific photoperiod then transferred to constant conditions, those with aberrant clocks will tend to emerge at unusual times. An experiment like this can be performed on a large scale with ease, using automated collection equipment (49): flies selected by the screen are collected and bred, and a small fraction of them can be expected to have a definable genetic lesion in a clock gene. This screen, or similar, was used to identify many of the major components of the circadian clock, for example, period (131) and timeless (207).
There is another lesson to be learned from this example. Although there is a massive literature on circadian clocks in humans, the human period-1 gene, on chromosome 17, was identified only in 1997 (220, 224), 10 years after the Drosophila sequence had been published (48). So, progress in human physiology might sometimes be accelerated by close attention to progress in simpler models.
4. The inositol trisphosphate receptor
Calcium signaling is a fundamental process, and in most cells, calcium is released from internal stores in the endoplasmic reticulum (ER) in response to hormone signaling, through the action of inositol 1,4,5-trisphosphate (IP3) on its receptor (IP3R), a large ER calcium release channel (24, 25). While most studies of calcium signaling have been performed on cultured cells, because of the experimental tractability, there have been some genetic experiments. In mouse, both the natural opisthotonos mutant and targeted knock-outs of the IP3R show multiple neurological defects and die soon after birth, confirming that, despite the existence of alternative ER release channels, the IP3R is essential for life. In Drosophila, the lethal phenotype is reproduced; however, here it has been possible to produce a range of alleles of differing severity (an "allelic series"), that includes severe hypomorphs, and lethal alleles that make viable, but hypomorphic, transheterozygotes. It is also relatively straightforward to construct chimeras so that lethal mutations can be studied in clones of homozygous cells in an otherwise viable heterozygote. In this way, slightly more informative results can be obtained; the key defect in hypomorphic alleles seems to be delayed larval development and adult emergence (101, 233). This implies a critical role for IP3R in the endocrine control of growth and development, probably connected to the release of the molting hormone ecdysone (233, 234). Intriguingly, although IP3R is single copy in the Drosophila genome, it does not seem to be necessary for visual phototransduction, a classical phospholipase C-mediated pathway (2, 182).
Such insights are not confined to studies in Drosophila. Similar analysis of IP3R function in C. elegans was taken in a slightly different direction by the generation of green fluorescent protein (GFP)-tagged IP3Rs, and by overexpressing the IP3 binding domain, to produce a dominant negative effect (19). The former technique provides a dynamic means of following levels and distributions of gene products in living animals and showed that, although IP3R is presumably pretty ubiquitously expressed, it is found preferentially, not just in the nervous system, but in the pharynx, gonad, intestine, and excretory cell (19). From here, it was relatively straightforward to map the gene's control regions and correlate them with the observed expression levels (90). The latter, dominant negative approach (together with RNA interference) showed a role for IP3R in two behaviors at opposite ends of the alimentary canal: feeding and defecation (235). The use of a heat-shock promoter for the dominant negative construct allowed intervention in the adult, avoiding the confounding effects of disrupting IP3R during development. These interventions would be difficult in a larger animal. It has also been shown in a yeast two-hybrid screen that IP3R interacts directly with myosin; the availability of a simple model allowed these results to be tested directly in C. elegans, confirming the functional significance of the interaction in vivo (236).
From the examples above, it can be seen that harnessing the power of even a simple genetic model can produce fundamental insights in mainstream (i.e., de facto mammalian) physiology. It is of interest to note, though, that the three examples shown above are all of the "brain and behavior" phenotypic class. There have been very few successful attempts to exploit Drosophila or other simple models in other areas of physiology. Later, this article will address perhaps the most developed such phenotype for epithelial function, ion transport, and cell signaling, the Drosophila Malpighian (renal) tubule. This will illustrate both that real physiology is possible and that it can provide useful general insights.
|
III. WHAT IS FUNCTIONAL GENOMICS? |
|---|
|
TopPreviousNextReferences |
|---|
|
IV. WHY GENOME PROJECTS NEED MODEL SYSTEMS |
|---|
|
TopPreviousNextReferences |
|---|
|
|
V. WHAT IS REVERSE GENETICS? |
|---|
|
TopPreviousNextReferences |
|---|
|
Perhaps less obviously, it further relies on the availability of a phenotype that can be studied for effects of the induced mutation. In other words, there is no point in mutagenizing a gene in a model organism in which there is no assay for effects of the mutation! For example, it would be a waste of effort to mutagenize a neural-specific gene in an organism too small to allow neurophysiological analysis. The choice of model is thus a trade-off between biomedical relevance, genetic power, and amenability to physiological analysis, exactly as we have outlined for integrative physiology above.
A. Reverse Genetics Demands a Genetic Model
Reverse genetics demands at the least some means of targeted mutagenesis of a gene of interest. Although generic approaches are being developed [RNAi (109) is an example], targeted mutagenesis has generally only been developed in organisms in which there is a considerable volume of classical genetics. Typically, this means that there is also a large body of literature covering known genes, visible markers, and recombination or physical maps. Such organisms have also invariably been selected as early subjects for genome projects. This confluence of classical and reverse genetics, together with genomics, provides a working definition of "genetic model" organism. Such organisms form a short but distinguished list and have impacted on nearly every field of biology to date, except physiology.
B. Even Human Genomics Requires Simple Models
Because reverse genetics will be essential to human functional genomics, it follows that model organisms are also essential. This explains the apparent contradiction that human genomics in particular, and biomedical science in general, has hugely cross-subsidized fundamental research in genetic models. However, as described in Table 1, the choice of transgenic model is not automatic. There is a trade-off between biomedical relevance (with humans top, and mammals next), cost, and convenience of life cycle (which favors smaller model organisms), ethical desirability (which favors smaller model organisms), and power of the genetic tools available (which favors smaller model organisms).
At first, mouse would seem to be ideal, as it is the only mammalian model for which detailed genetic intervention is routine. It is possible to inactivate genes with single base-pair precision by homologous recombination and to introduce transgenes. However, gene knock-outs are often lethal sufficiently early in development as to be uninformative. In mouse, the ability to introduce more subtle defects (for example, to inactivate gene function only in adults, or only in a particular tissue) is highly limited and must be set up on a case-by-case basis, compared with other models, like Drosophila or C. elegans, where generic technologies are available. Accordingly, for the foreseeable future, these organisms are likely to be ideal models for integrative physiology.
C. Other Routes to Functional Analysis
Although it is generally accepted that reverse genetics is the most powerful method of elucidating function of novel genes, there are other special cases that are nonetheless worth noting. For example, cluster analysis of microarray data tends to gather together functionally related genes. So if a single novel gene clusters with known genes that are all implicated in a single process (like amino acid metabolism), then a clear functional hypothesis presents itself (76). Similar routes can be plotted at the proteome level, although there are more potent ways of identifying functional complexes. By systematically fishing out the partners of epitope-tagged proteins in yeast, it has proven possible to identify the constituents of multiple hetero-multimeric complexes (86, 105). Similarly, more advanced computational techniques may allow hypothetical proteins to be classified based on structural similarities more subtle than conventional BLAST searching at present (77, 111, 175). However, these exceptions also illustrate the rule: they all generate hypotheses that must be tested experimentally. So reverse genetics is still a required part of the transition from gene to function, and physiologists are essential participants in the endeavor.
|
VI. THE "PHENOTYPE GAP" |
|---|
|
TopPreviousNextReferences |
|---|
It is easy to demonstrate the phenotype gap objectively. For example, most rodent physiologists have concentrated on rats, guinea pigs, and rabbits rather than the smaller genetic model, the mouse. This difference can be marked: a literature search of Science Citation Index publications shows that for almost any keyword or tissue of choice, the recent literature on rats is about fivefold larger than on mouse (Table 2).
|
Although rat genetics and genomics will eventually rival those of the mouse in power, there is a pressing need to translate rat physiological methodology into mouse, and so provide the widest range of physiological assays possible. This will maximize the chances that mutation of any particular gene will have a detectable and informative physiological phenotype.
Genome projects acknowledge the need for physiologists to help close this phenotype gap, and this provides some of the most exciting openings for physiologists into the 21st century. Exactly how to ride this wave is discussed later.
In the next section, we will give an example that draws together the strands of the article: that combining physiology and genetics in a simple genetic model provides an ideal system for truly integrative physiology, and that by doing so physiologists can contribute to post-genomics by closing the phenotype gap.
|
VII. THE DROSOPHILA MELANOGASTER MALPIGHIAN TUBULE |
|---|
|
TopPreviousNextReferences |
|---|
2 mm long by 35 µm in diameter, and are comprised of 150 cells. We have recently found that the Drosophila renal tubule can be used for physiological studies of fluid secretion (69) and that it shows a rich pharmacological repertoire that includes the cAMP, nitric oxide (NO)/cGMP, and Ca2+ signaling pathways (63, 66). This means that, for a very wide range of developmental, transport, and signaling genes, the tubule is probably the most robust Drosophila phenotype presently available. A. History of Malpighian Tubule Physiology
This has been reviewed briefly elsewhere (63). Marcelo Malpighi (1628–1694), physician to the Pope, is well known as an early adopter of the microscope, contemporaneously with Hooke, and followed the discoveries of the great William Harvey on the nature of circulation. However, he was a comparative anatomist and also turned his microscope to insects, discovering the eponymous insect Malpighian (renal) tubule. However, it was not until the 20th century that insect physiology became feasible. Wigglesworth, then his pupil Ramsay (183), was able to demonstrate that insect tubules indeed produced urine, and Maddrell (150) was able to demonstrate endocrine control, using one of the more remarkable epithelial model systems, the tubule of the bloodsucking bug Rhodnius prolixus. As this bug feeds on huge blood meals perhaps once in 6 mo, and must then fit into cracks in walls and floorboards to escape discovery, it has a potent control of its diuresis and can accelerate urine production by more than a 1,000-fold after feeding (153). Since then, there has been a large literature on insect tubules from many species (for reviews, see Refs. 50, 62, 63, 66, 150, 152, 153, 173, 177, 231), and it is accepted that the insect tubule is both a useful epithelial system and a valid target for insecticide development. This classical literature suggested that the Malpighian tubule of Drosophila melanogaster might be amenable to physiological analysis, although it was smaller than any previously studied.
B. How Is the Drosophila Malpighian Tubule Organized?
Insect Malpighian tubules are simple epithelia, free-floating in the insect's hemocoel (there is no vascular circulatory system beyond a rudimentary tubular heart) (43, 247). Similarly to birds, and for similar reasons of water economy, the tubules do not open directly to the outside, but join to the alimentary canal at the junction of the endodermal midgut and the ectodermal hindgut (Fig. 4). The relative experimental accessibility of the insect tubule compared with the vertebrate kidney tubule is an advantage; they are similar in overall dimensions. In the case of Drosophila, there are four tubules, joined in pairs through short common ureters to the alimentary canal. Interestingly, the right-hand pair are longer, with a prominent white initial segment, and are always placed anteriorly in the body cavity. In contrast, the left-hand pair was thought to lack initial or transitional segments, and always sit posteriorly, within the abdomen (242). (Incidentally, this could provide a useful screen for mutants of sinistrality, an area hardly explored in fly). There are two major cell types: the larger type I (or principal) cell and smaller intercalated type II (or stellate) cell (242).
|
2. Enhancer trapping as a tool to understanding organization
A fundamental part of a physiological study is to analyze the structural organization, or functional morphology, of the tissue of interest. In a nonmodel organism, our understanding of a tissue is based on the experimental techniques available and the training of the scientist. It is thus only possible to obtain an experimenter's view of the tissue organization. In certain genetic model organisms, in contrast, a technique known as enhancer trapping can provide an insight into how the organism organizes a target tissue.
3. The principle of enhancer trapping
Enhancer trapping has been widely used in the context of development, for which it was originally developed (23). It makes use of an engineered transposon, in Drosophila usually the P-element (193). Transposons are a class of semi-autonomous mobile DNA elements (loosely similar to retroviruses), found widely in all organisms, that encode an enzyme (transposase) that can recognize the stereotyped ends of the transposon, and catalyze its excision and reinsertion in the genome (187). In most organisms, this parasitic process is largely benign, although occasionally a cell can be killed or transformed by disruption of a key gene ("insertional mutagenesis"). In Drosophila, one transposon (the P-element) has been engineered to carry a variety of genetic constructs that permit a remarkable range of genetic manipulations in the intact organism (192). For enhancer trapping, a P-element is used in which the transposase gene has been replaced by three elements: 1) a white marker gene, which gives red eyes in flies carrying the transposon and so allows it to be tracked through crossing schemes; 2) a reporter gene (such as lacZ), coupled to a weak, permissive promoter [expression of the lacZ gene is then sensitive to its genomic context (i.e., where the transposon is sitting within the genome, relative to other genes and their natural promoters and enhancers)]; and 3) a linearized plasmid, which allows the site of insertion to be determined by a technique known as plasmid rescue.
Without a transposase gene, this element is capable of mobilization but cannot transpose itself. It is thus trapped within the genome unless provided with a source of transposase. This enzyme can be supplied by crossing flies carrying the P-element to another line in which a P-element has been serendipitously inactivated by an imprecise excision event in the past. In the 
2,3 line, a functional transposase is produced only in the fly's germ cells, and as the element is missing one of its ends, it cannot be mobilized by its own enzyme (187). In an enhancer trap screen (Fig. 5), flies carrying both the enhancer trap element and a source of transposase are called jump-starters. In each of their sex cells (usually sperm: males are used by preference because there is no male meiotic recombination), there is the potential for the enhancer trap element to mobilize and reinsert somewhere else in the genome. By segregating and breeding true from hundreds or thousands of progeny of such jump-starter males (and by ensuring that the 2,3 transposase source is quickly crossed out of each line), a range of new, stable, insertion sites will be obtained. In some of these, the enhancer detector will have landed near some unknown gene with patterned expression in the tissue of interest, so by staining representatives of each line for the reporter gene, the 10% or so of genes of interest to a particular experimenter can easily be determined. (Of course, the remaining lines may be of interest to another experimenter, and it is common practice to exchange panels of lines within the Drosophila
community.)
|
This description may sound esoteric in a physiological context, but its execution is straightforward, requiring only a series of crosses over months, and the physiological payback is huge, because the technique reveals the organism's, rather than the experimenter's, view of the organization of the tissue. In contrast to the methodology, the results are visually striking and easily grasped.
4. Enhancer trapping can detect tubule-specific genes
As part of a 1,500-line screen using the P{GATB} transposon (252), 750 lines were screened for tubulespecific patterns of expression in larvae and adults (214). Of these, 10% were informative, and of these, 20 lines were used to delineate boundaries of expression in the tubule. In most studies, enhancer trapping is used as a rapid means to identify genes of interest, and so generate data similar to the results of a differential cDNA library screen. However, the patterns themselves are informative, as they reflect the actions of combinations of cell-specific transcription factors on the enhancer detector. In this way, they genuinely reveal aspects of the tissue's spatial and temporal organization.
5. Enhancer trapping reveals regional specialization
Classically, the anterior tubules were considered to have three domains and the posterior tubules one. We found lines that reflected these domains of expression, consistent with the previous studies (Fig. 6, A—D); however, we also found new aspects of organization. The lines that marked out initial and transitional segments in the anterior tubule also identified miniature domains in the posterior tubule (Fig. 6, A and B). The anterior and posterior tubules thus differ quantitatively in the extent of these regions, rather than qualitatively in the nature of their specification. The main segment of the tubule (Fig. 6C) can be divided into two, with a prominent lower tubule domain (Fig. 6D) that is in turn separable into three domains (214). Critically, then, there are six regions in both anterior and posterior tubules, and several of them had been refractory to experimental identification in the absence of enhancer trap studies.
|
6. Enhancer trapping reveals multiple cell types
The principle of these studies can be extended to cell types (Fig. 6, E—H). Surprisingly, few lines marked out all the principal cells in the tubule: the normal pattern was for only a subset of principal cells to be marked (Fig. 6E). This is significant because it means that morphologically indistinguishable cells are expressing different genes, and thus presumably performing different functions.
The other major cell type, the stellate cell, is also clearly labeled by this technique (Fig. 6F). Interestingly, the two lines that marked stellate cells had the most restricted pattern of expression in the rest of the fly of any studied. This may imply that their function is unique and that they express relatively unusual transcripts. The distribution of stellate cells also helps to cross-validate the domains of expression outlined earlier: stellate cells are not found in the lower tubule domain, and their shape changes from stellate in the main segment to bar-shaped in the initial and transitional segments (Fig. 6G). The bar-shaped cells can in turn be distinguished from stellate cells by a single line that labels only the former.
There are other, minor cell types, such as tracheal cells, and tiny myoendocrine cells found only in the lower tubule region (Fig. 6H). Taken together then, there are six cell types and six regions in this tiny tissue.
7. Genetic domains can be quantified by counting nuclei
At this stage, it would be traditional to clone the flanking genomic regions by plasmid rescue and seek to identify the genes responsible for the patterns of expression that had been observed. However, we tried to quantify the sizes of different domains and to compare the sizes and stability of domains reported by different lines to see whether they were reporting the same boundaries. This would render the description of tubule structure at a quantitative rather than anecdotal level. In fact, it was straightforward to label tubule nuclei with ethidium bromide and to count the nuclei in each domain (Fig. 7A).
|
Remarkably, it was possible to show that the positions of cell boundaries, and the numbers of cells in each region, were fixed to near single-cell precision (Fig. 7B). This means that every cell in the tubule has a precise view of its positional identity, and the development of the tubule is extremely robust and deterministic. Although all the insertions studied were homozygous viable, some are nonetheless disruptants of the genes in which they were inserted (see below). Despite this, cell numbers did not vary between lines, in any of those we studied, although, of course, there are some developmental mutants in which tubule development is perturbed (106). In this simple, one-dimensional system, it may be possible to undertake identified-cell epithelial physiology, for almost the first time. Although this is already technically possible for morphologically distinguishable cells (for example, Refs. 30, 95), it is clear that morphologically indistinguishable cells can express different genes (Fig. 6), and so have different functions (151, 171, 186). Enhancer trapping, and particularly the vital labeling of domains of gene expression, neatly provides a solution to this issue.
The level of sophistication revealed by enhancer trapping might seem daunting, but the same enhancer trap technology provides the tool to intervene genetically in any population of cells that can be defined by an enhancer trap line.
In the second generation, or binary, enhancer trap system (Fig. 8), the reporter gene is the yeast transcription factor GAL4. This appears to be almost perfectly inert within the Drosophila genome, and so normally has no effect on the host organism. However, it is capable of driving transgenes under control of the yeast UASG promoter. Thus, given a panel of informative GAL4 lines, as described above, any genetic construct of choice (not just lacZ) can be directed to any of the lines. The GAL4/UAS system additionally acts as a switch, in that high levels of driven expression contrast with very low background levels. The binary nature of the system means that a new transgenic line can be generated in 3 mo, without having to repeat the enhancer trap search for GAL4 drivers. This compares very favorably, in both time and cost, with the effort to make knock-in constructs for mouse and to generate and test the transgenics. There is another key advantage of the system: if a deleterious gene product (for example, a gene in the apoptosis pathway) is to be expressed, then the UAS stock can be kept safely without serious loss of fitness. The deleterious construct is only expressed in the progeny of the cross between the parent lines.
|
This GAL4/UAS binary system opens possibilities to the physiologist that are mouth-watering. At present, it is a technique specific to Drosophila, and a potent reason why experimenters should relish, rather than dread, the opportunity to work in this organism.
9. Correlating functional and genetic maps
Enhancer trapping suggested a clear, reproducible structure for the Malpighian tubule. However, it is critical to establish whether this genetic map has any relevance to physiology or is merely some genetic curiosity. To accomplish this, the genetic map was tested with a battery of functional assays, and in each case functional and expression domains were found to coincide perfectly. This section reviews evidence for the congruence between genetic and functional maps.
A) FLUID SECRETION. The key property of the Malpighian tubule is its ability to produce an isotonic fluid. Indeed, for most insects, this is the only property that has been measured. It is possible to adopt the classical Ramsay assay for fluid secretion to this tiny tissue, and the Drosophila tubule is actually very robust ex vivo, maintaining stable secretion rates of 0. 5–1 nl/min for several hours in appropriate medium (69). It is possible to map secretion rates as a function of length along the tubule, although the spatial resolution of the assay is not as good as that of enhancer trapping. However, it was possible to show that the main segment of the tubule secretes fluid, while the lower tubule reabsorbs it (Fig. 9) (171).
|
B) ALKALINE PHOSPHATASE. A powerful product of the enhancer trap technology is the rapid identification of the genomic location of the insertion and thus frequently the gene near which the P-element is inserted. The lower tubule is delineated by two lines, which represent independent insertions in the same gene, encoding an alkaline phosphatase (Aph4) with greatest similarity to the human liver/bone/kidney (ALPL) type (253). In humans, ALPL is an ecto-enzyme that metabolizes phosphoethanolamine (PEA) and pyridoxal-5'-phosphate (PLP). In Drosophila, Aph4 is expressed only in the lower domain of the Malpighian tubule, and in a small group of cells, the ellipsoid body in the brain (253). It also appears to be an ecto-enzyme, as histochemistry for alkaline phosphatase activity labels only the apical surface of the lower tubule.
Remarkably, although the genome project has annotated 13 alkaline phosphatase genes in Drosophila, the expression patterns of the Aph4 enhancer trap insertions perfectly match the alkaline phosphatase histochemistry. This implies either that the other 12 alkaline phosphatase genes are not expressed significantly in the adult or that the nitro blue tetrazolium-based histochemical stain does not detect all the alkaline phosphatase activities in Drosophila.
In vertebrates, the roles of alkaline phosphatase are not entirely clear. In humans, mutations in ALPL are associated with hypophosphatasia (241). In mouse, the homologous tissue nonspecific (TNAP) isoform, when mutated, causes fatal seizures 2 wk after birth; these can be survived if pyridoxal is administered. There is also some evidence of hypomineralization in teeth (240). At least one human case has been reported with similar symptoms (244). However, the enigmatic renal role of alkaline phosphatase might be addressed in Drosophila, as the P-element insertions in Aph4 both disrupt expression and cause a transport phenotype (Fig. 10).
|
10. Phylogenetic scope of enhancer trapping technology
It is clear from the above that enhancer trapping delivers results that are highly relevant to the physiologist at several levels. First, it provides a screen for tissue-specific genes with added informational content by virtue of the expression pattern reported. Second, enhancer traps themselves report pattern in a tissue at a level that could probably never be deduced by explicit experimentation, so delivering new understanding of tissue organization. Third, second and later generation enhancer traps provide generic technologies for conditional expression, overexpression, and gene tagging. Fourth, enhancer trap insertions have the potential to be mutagenic, either directly by virtue of their insertion site or after imprecise excision of the P-element. They are thus vital components of a functional genomics strategy (22).
Obviously, these experiments can never be attempted in humans, so it is necessary for human physiologists to resort to model organisms. Enhancer trapping is at its most developed in the fly, with multiple classes of P-elements available, including the second generation GAL4/UAS binary system, together with several ES and other gene-trap vectors (22, 189).
In Arabidopsis, enhancer trapping and gene trapping have both been used, usually with the maize Ds transposon and GUS as the reporter gene (129, 221). However, the technology does not appear to have been used to its full potential (22).
In zebrafish, a Rous-sarcoma virus LacZ reporter injected into embryos showed patterned expression, although in only one line was it heritable (18). This is a fairly laborious method, as each line is derived from a separate injected embryo; there is no remobilization of a preexisting integrated reporter gene.
In C. elegans, transposable elements are not easily tamed, because of the lack of clean germ line-specific transposase expression (22). However, it is relatively easy to introduce heterologous DNA into worms, and this has been used for a form of promoter trapping in which random genomic fragments are cloned upstream of a LacZ reporter, then transformed into worms (145). If this technique is performed systematically on a large enough scale, it can emulate some of the functionality of enhancer trapping.
Enhancer trapping is technically possible in mouse, although the large genome size means that fewer insertions are likely to be informative. Additionally, the effort in generating and maintaining sufficient stocks (the limiting factor in Drosophila screens) makes a rigorous enhancer trap screen unlikely. However, promoter and gene-trap techniques have been applied widely (191). Such studies usually rely on a splice acceptor site within a retrovirus vector, producing a lacZ fusion protein with the initial exons of the mouse gene. Such insertions are both rarer and much more mutagenic than enhancer traps, because they are guaranteed to disrupt the protein. Rather than try to jump the vector within the mouse, random insertions are generated within clones of embryonic stem (ES) cells. These can then be grown up into whole mice as desired. This can produce a valuable, high-throughput screen, as insertions can be screened for expression, or the insertion site sequenced, within ES cells; transgenic mouse lines are then only generated from cell lines of interest (116). At one level, this has less appeal than Drosophila enhancer trapping, because it implies preselection by the experimenter, and so reintroduces experimenter bias into the paradigm. However, at another level, it provides a potentially valuable route to reverse genetics of defined genes, perhaps those identified from high-throughput methodologies. Such an approach has been taken to a highly commercial level, with experimenters able to purchase ES clones with defined insertions from a bank of over 200,000 frozen lines (254) and produce their own transgenic mice (http://www.lexgen.com/omnibank/omnibank.htm). There are obvious cost and intellectual property issues in such an approach.
Elements of the GAL4/UAS system are now used in mouse, in order specifically to provide switched expression (138). However, in mouse, GAL4 cannot be
