Biotechnology and the Integration of Behavioral Science:

A Report to the Division of Integrative Biology and Neurosciences, the National Science Foundation.

Prepared by
Robert E. Page, Jr.
Department of Entomology
University of California
Davis, CA 95616

and

Robert Hitzemann
Department of Psychiatry
State University of New York
at Stony Brook
Stony Brook, NY 11794

 

 

This document summarizes the results of a workshop held July 6-8, 1999 at the National Science Foundation in Arlington, VA. The workshop, "Biotechnology and the Integration of Behavioral Science," was organized to bring together researchers working in different areas of behavioral genetics, using different analytical tools, in order to explore ways in which the National Science Foundation can continue to promote the integration of animal behavior, behavioral genetics, and neuroscience. The workshop was supported jointly by the Animal Behavior and Behavioral Neuroscience Programs in the Division of Integrative Biology and Neurosciences.

Great technological advances recently have been made in molecular biology that have greatly impacted these fields to different degrees and provided different tools. These advances have also set these subdisciplines on different trajectories. It is the belief of the organizers of this meeting that intellectual and informational bridges must be maintained in order to keep an integrated behavioral science. Without them, behavioral science is in danger fragmenting into subdisciplines based on the techniques employed resulting in the loss of focus on animal behavior. In addition, researchers who are primarily interested in explaining variation in behavior at the level of the organism interacting with its environment should be knowledgeable in behavioral genetics. This includes those who have a primary interest is the evolution of behavioral change — behavioral ecologists. This critical information sharing will not take place if the field of animal behavior does not become better integrated.

Why Study Behavior?

Behavior lies at the interface of an organism and its environment. This is the domain of natural selection. Behavior is the result of different levels of biological organization acting in concert: genes, cellular physiology, neuronal anatomy, skeletal and muscular anatomy, etc. Changes in behavior occur in different time scales. Short term changes occur during the life of the organism result in behavioral plasticity while long term changes occur on an evolutionary time scale resulting from natural selection on heritable variation. So, behavior is the culmination of the integration of a rich set of biological processes. But, above all, only by understanding behavior can we understand ourselves and our place in the natural world.

Objectives of the Workshop

The main objective of this workshop was to explore how new emerging technologies, for example gene mapping, functional genomics, bioinformatics, and gene transformation, can be applied to a better understanding of animal behavior at all levels. The relationships between behavioral genetics, neurobiology, ethology, and behavioral ecology, are obvious. Collectively they encompass behavioral science. Our challenge was to determine ways to integrate them. We brought together researchers working at these different levels (including the gene, physiological, organismal, social, and ecological) to promote the exploration of integrative approaches to the analysis of behavior and underscore its importance.

We believe that those objectives were met. It was apparent from the beginning that we as a group represented different subdisciplines within animal behavior and many of us had little knowledge, understanding, or appreciation for one another’s approaches. Everyone felt enriched by the interaction at the end of the three-day workshop. We saw a unification of approaches around common research objectives and stimulated interest in each other's work. The workshop provided and appreciation of the breadth of behavioral science and resulted in establishing collaborative projects between different participants who study behavior and different levels of organization. It was apparent the meeting was needed, and successful.

Structure of the Meeting

A total of 18 participants provided 30 minute presentations of their work. Sessions were organized according to broad themes, and discussions were held after each session. The final day was devoted primarily to discussing broader issues designed to develop specific recommendations for NSF administrators. Participants, titles of their presentations, and discussion questions may be found in the attached appendix.

 

General Summary of Discussions

Common objectives

Aside from the objectives of the workshop, there are objectives for behavioral genetic research. Many of these objectives are common to all approaches:

1. Determine the number of variable genes, and their relative contributions, that affect observed variation in behavior.

2. Understand gene function -- the physiology and biochemistry of the gene. How do genes interact? This includes allelic dominance, epistasis, and pleiotropy.

3. Understand the evolution of behavioral traits. This encompasses understanding the amount and nature of naturally occurring genetic variation; the roles of genes, environment, and their interactions in determining behavioral phenotypes; comparative behavior and genomics; the evolution of DNA sequence and how it relates to behavioral evolution; and understanding the phylogeny of behavior.

Forward genetic approaches

Approaches to understand the genetic basis of behavior can be broadly categorized in into forward and reverse genetics. The forward genetic approach starts with identifying variation in a behavioral phenotype then attempting to understand the underlying genetic basis of that variation. Reverse genetics starts by discovering genes, then determining how they affect phenotypes. Again, many of the objectives are the same, but some differ, but they do differ fundamentally in their approaches.

Quantitative genetics

Quantitative genetic approaches are forward. Observed variation in behavioral phenotypes is partitioned statistically into variance components that correspond to genetic and environmental effects on the phenotype. As a research tool, it is particularly well suited for this task and can reveal the complex, interactive, genetic structure underlying behavior. It is an excellent tool for studying natural populations and nonmodel organisms because a genetic information can be extracted with relative ease and minimal technological burden. Comparative evolutionary studies are facilitated by the relative ease of experimental design and analysis.

Quantitative genetic approaches treat the phenotype as the resulting phenomenon and attempt to explain phenotypic variation on the basis of genetic "factors", rather than attempting to identify the genes themselves. It is often the logical, and very important first step toward understanding the nature of genetic variation for behavior but is much more effective in combination with QTL and functional genomics approaches that may lead ultimately to the genes themselves.

QTL mapping

This approach grew out of quantitative genetics and addresses the same fundamental question: what is the nature of quantitative trait variation? It differs from quantitative genetics in that it is focused on the genome. QTL mapping employs molecular biology and sophisticated computational algorithms as tools to dissect the genome and isolate genes responsible for observed behavioral variation. It has proven to be a powerful tool for determining how many genes affect traits and their complex interactions with each other and the environment. There are three stages of QTL analysis:

1. Detection -- The detection of a QTL can be done with classical quantitative genetic approaches. In that case, you can determine that there are QTLs somewhere in the genome that are affecting a trait of interest, such as a specific behavioral phenotype. You may even determine that the gene is linked with some other gene or marker on a given chromosome.

2. Mapping -- A genomic map is constructed using genes or markers that identify linkage associations. Classical breeding methods are often employed such as the F1 backcross or F2 intercross. Many new tools have recently been developed to assist in map construction. These include better methods for expressing variable genetic markers, such as RAPDs, AFLPs, and microsatellites, and better statistical methods for map construction. This makes it possible to rapidly construct linkage maps even for non-model organisms.

3. Fine mapping -- Stage 2 can locate a QTL within a region of a chromosome but will not be sufficient for gene discovery, identification, and characterization. A more detailed map must be constructed to isolate the QTL to a region small enough to begin a search for candidate genes. Special marker saturation techniques such as bulk-segregation analysis and genome walking, and special breeding designs such as advanced intercrossed lines, heterogeneous stocks, and selective phenotyping are necessary.

Currently, the field of QTL mapping is in Stage 3. Many fine scaled maps are being produced that reduce the amount of DNA sequence that must be searched for functional genes. The next steps will be to actually identify genes with major effects through positional cloning. Genes with minor effects are more difficult to detect but should eventually be within the capability of some QTL mapping projects. Linked QTLs continue to be difficult to separate but should be possible with improved computational models, highly saturated linkage maps, and large data sets. Many of the details of gene-gene and gene-environment interactions should also be possible through QTL mapping.

The biggest challenge facing the QTL approach is demonstrating that you can actually find the genes that are mapped as QTLs. So far, there are no examples of successfully going from phenotype to isolating the responsible gene(s) using this approach. Until that happens, the QTL approach, like quantitative genetics, will appear to fall short of the ultimate objective, finding the responsible genes. However, well-saturated QTL maps of well-defined behavior with tight positional confidence limits will be a tremendous asset within the next few years when more genome information is available. At that time localization and isolation of genes on the basis of QTL maps will become routine.

By its nature the QTL approach is designed to dissect multigenic traits. Therefore, it depends on "natural" variation in traits, so, only naturally varying genes are detected. Many genes are fixed by natural selection, therefore, they are invisible to this technique. Fixed genes, however, may have been very important in the evolution of the trait. Relying on natural variation will not work in those cases.

Gene discovery using QTL mapping approaches will require verification. Verification will require the use of different tools such as mutagenesis and transformation. These tools will not be readily available for many of the organisms studied. New methods must be developed for QTL verification in non-model organisms. One promising approach is down regulating genes using fragments of RNA.

Reverse genetic approaches

Most objectives are shared by forward and reverse genetic approaches. However reverse genetics has some objectives that are not, yet, in the central domain of forward genetics. These primarily involve the function of genes, functional genomics. This difference in the domains of these approaches is only temporary and represents an historical constraint on where different researchers started in their search for genes underlying the behavior they study: Objectives are:

1. Understanding changes in gene expression during development

2. Mapping the spatial and temporal patterns of gene expression associated with behavior, in particular learning and memory

3. Defining specific gene-gene-environment interactions

4. Characterizing genes at the molecular level

5. Understanding the behaviorally relevant cellular function of genes

One approach is to identify a variable gene then determine its effects by studying natural populations. However, few naturally occurring single gene behavioral traits have been identified. More often, mutagens are used to induce allelic variation. Behavioral phenotypes are screened and variation is identified and mapped back onto the genome to identify the gene responsible. This allows the study of the function of fixed genes. Another approach is to study in novel organisms the effects of genes that have been characterized in model organisms, facilitating comparative genomic studies. In situ hybridization studies of gene expression in brains of novel organisms using sequence information obtained from other organisms, is an example.

Reverse genetics currently employs a different set of tools from forward genetic approaches. Reverse genetics relies heavily on mutagenesis, target knockouts, expressed sequence tags (EST), antisense knockdowns, and in situ hybridization. Techniques on the horizon that are destined to facilitate this approach include EST microarray chips, RNA-i regulation of gene expression, and viral delivery systems for pharmacological agents.

Like the other approaches, this one is not without difficulties. The main problem with mutagenesis and knockouts is that they are "blunt tools". Genes may have multiple functions at different times during development or in different behavioral contexts. The complete elimination of the function of a gene may be lethal or so disruptive that its function in a specific behavioral context is masked. As a consequence, there is a need to be able to control the time and place of expression of genes, something which is not yet readily attainable.

Functional genomic studies using microarray chip technology can produce enormous amounts of data relatively quickly, at a high financial cost. However, the enormous number of "discovered" genes that appear to vary in expression must be independently and individually verified. This could be a monumental task. In addition, there are no sufficient methods for determining which apparent differences are statistical artifacts and which are real. Type I and type II statistical errors are likely to be large. The "real story" of the effects of genes discovered using this technology will probably reside in their interactions, a level of complexity far greater than finding and verifying individual genes. Will we be able to construct an understanding of the whole from studying so many interactive parts?

The "dichotomy" between reverse and forward genetics is not rigid, and probably only temporary. Mutagenesis has been used as a forward genetic tool in studying learning and memory, and ESTs have been used to define potential genes responsible for observed differences in learning phenotypes. Mutagenesis and targeted knockouts have been used to verify genes discovered as QTLs using forward genetic techniques. In the future, these tools will be used interchangeably.

Consensus Summary

Although there was not unanimity of opinion, the consensus view of the participants at the workshop is that the different behavioral genetic approaches presented are complementary. The combination of quantitative genetics, QTL mapping, functional genomics, bioinformatics, and neurobiology provides the best tool kit for studying the relationships between genes and behavior. Researchers should be willing to employ the best tools for their particular objective. This workshop gave each of us a greater appreciation for the other's approach and unified these approaches around common objectives.

Recommendations

The following is a set of recommendations that were derived from a specific set of questions discussed by the participants at the conclusion of the workshop. These are very general in nature. We realize that there may already be some programs within NSF that meet these needs.

What research resources are needed to promote molecular genetic studies of behavior in non-model organisms?

1. Funding for developing techniques

The perception is that it is very difficult to get funding through normal NSF program channels to develop the technology needed to address new and important questions in behavioral genetics. Most of the technology is borrowed from model organism research, developed with funding from NIH. However, the needs of researchers studying interesting behavior of non-model organisms differs from those working with model systems.

2. Money for research resources

Research resources are difficult to maintain without long term consistent levels of funding. Too often, resources are lost due to the 3 year funding cycles at NSF. Behavioral genetic research as described above requires a commitment to resources over the long term. This research cannot be maintained with stop and go 3 year funding which is the norm for NSF. We especially need mechanisms for the following:

a. Developing and maintaining large data bases -- Large, shared data bases are the way of the future. Laboratories need to combine and share their behavioral and genetic information.

b. Build sequence databases -- Sequence databases, such as sequence from expressed genes, will be the key to functional genomics and gene discovery. Currently, a few model organisms have been targeted by NIH for establishing sequence data bases. NIH should get involved with a focus on developing data bases for exploring interesting behavior.

c. Establish cDNA and genomic libraries as data resources -- Gene libraries go hand in hand with sequence data bases. It is very difficult to get funding within a NSF program to develop a cDNA library, or a BAC library to use as a resource for gene hunting. But, it is essential that those resources be available.

d. Fund development of exploratory technologies -- Some technological advances have the potential of having a huge, broad impact on the field of behavioral genetics. NSF should promote and fund multi-lab projects designed to develop and/or test new promising technology. One example is RNA-i technology for regulating gene expression.

e. Maintaining genetic stocks of nonmodel organisms -- Many researchers in behavioral genetics work with nonmodel organisms that require resources to maintain. Some have relatively long generation times, compared with drosophila or mice, and cannot be put in diapause when not needed. These stocks may require years to develop. But, with the on and off funding cycles of NSF these stocks may be lost because there is no current "hypothesis driven" funding available.

f. Instrumentation -- More money is needed for funding the purchase of instrumentation. These costs continue to increase, and instruments are often the first items to be cut from budgets of funded projects.

g. Technical personnel -- Some mechanism is needed to fund trained technicians. It takes years to train these people and they exist on soft money from grant to grant. Again, the on again-off funding patterns of NSF programs make it impossible to maintain good, well trained personnel. We need longer grants, or ways to maintain personnel between grants.

 

3. What levels of funding support are needed for gene-based behavioral research ?

This question was on our list. However, we did not feel that this was an appropriate question for this forum.

4. How can NSF promote a better exchange of information and technology between the disciplines of genomics, neuroscience, and animal behavior?

a. Keep this group together -- We believe that the participants in this workshop represent a unique combination of technological approaches, levels of investigation, and perspectives on animal behavior. There was an interesting synergism that emerged during the three days we worked together. We could constitute the core from which the objectives of this meeting may eventually emerge. This workshop was the beginning, it resulted in us getting acquainted with each other and the different areas of behavioral genetics represented. We need to have another workshop where we deal primarily with working out ways to implement the recommendations discussed here.

b. Encourage and provide support for cross attendance at professional meetings -- We came from different subdisciplines with behavioral genetics, and found that we often knew little about others were doing. That discipline blindness is even more pervasive in the animal behavior community at large. We need to find a way to go to each others professional meetings and gain a broader view of animal behavior and to educate those in other subdisciplines about what we do. This needs to be an active outreach program.

c. Train molecular biologists in behavior. It is clear that the future of behavioral genetics and animal behavior in general, like all biology, will be centered around molecular biology. There are many trained molecular biologists looking to retool and many of them find behavior fascinating. We need a postdoctoral fellowship program designed to retrain molecular biologists in behavior. The objective would be to also cross train a P.I. working in animal behavior in molecular biology.

d. Workshops and courses going from phenotype to QTL to gene, emphasizing technology training -- NSF should encourage and sponsor special workshops and training courses that deal with the technology discussed above.

 

D. Should NSF support a special initiative within the Biology Program directed at understanding the genetic architecture of variation in behavior? Would it attract sufficient proposals?

The unanimous opinion of the participants was -- Yes and Yes.

E. What is a unique roll for NSF?

We believe that NSF needs to delineate its roll relative to NIH. NIH is firmly focussed on human welfare, promoting the study of model organisms as surrogates. NSF can play a unique roll biasing its support for analyses of naturally occurring behavior. Studies with evolutionary perspectives, and especially comparative studies would be clearly within this purview. NSF should resist establishing a new set of model organisms, but instead should focus on interesting behavior.

Appendix

Biotechnology and the Integration of Behavioral Science

NSF Workshop

Organizers: Rob Page and Bob Hitzemann

July 6-8, 1999

Program

 

July 6, 1999

9:00 - 9:30 Introduction to workshop (Bob Hitzemann and Rob Page)

Mapping QTLs: Models and Theory (moderator: Bob Hitzemann)

9:30 - 10:00 Ariel Darvasi "The theory and practice of mapping QTLs."

10:00 - 10:30 Zhao-Bang Zeng "The genetic architecture of quantitative traits."

10:30-11:00 Break

11:00 - 11:30 Jim Cheverud "Epistatic interactions and complex genetic architecture."

11:30 - 1:00 Lunch

Mapping Behavioral QTLs: From Behavior to Genes (moderator: Rob Page)

1:00 - 1:30 Marla Sokolowski "Forward genetics and the foraging gene of Drosophila."

1:30 - 2:00 Robert Page "QTL analysis of the foraging behavior of honey bees."

2:00 - 2:30 Jeanne Wehner "QTL analysis of fear and anxiety related phenotypes."

2:30 - 3:00 Break

3:00 - 3:30 Bob Hitzemann "Detection, mapping, and fine mapping of murine behavioral phenotypes."

3:30 - 4:00 Clinton Kilts "Genetics of latent inhibition."

4:00 - 4:30 Jack Werren "Linkage mapping and the analysis of epistatic interactions in a parasitic wasp."

4:30 - 5:00 Greg Hunt "QTL mapping defensive behavior in honey bees."

July 7, 1999

8:30 - 10:30 Discussion (Moderators: Darvasi and Wehner)

Questions:

1. Can we really achieve an understanding of the genetic architecture of naturally-occurring variation in behavior using a forward genetic approach?

2. Can we really achieve an understanding of the genetic architecture of naturally-occurring variation in behavior using a reverse genetic approach?

Functional Genomics and Behavior (moderator: Rob Page)

10:30 - 11:00 Josh Dubnau "Functional and dysfunctional genomics of learning and memory."

11:00 - 11:30 Gene Robinson "Gene discovery and behavioral development in honey bees."

11:30 - 1:00 Lunch

Molecular Neurobiology (moderator: Bob Hitzemann)

1:00 - 1:30 Barbara Taylor "Genetics of sexual behavior."

1:30 - 2:00 Jochen Erber "Learning and memory in the honey bee"

2:00 - 2:30 Tim Tully "Genetics of learning and memory in Drosophila."

2:30 - 4:30 Discussion (Moderators: Robinson and Tully)

Questions:

1. Can functional genetic approaches lead to an understanding of the genetic architecture of complex behavioral traits?

2. Can functional genetics help us understand the relationships between brain function and behavior?

Classical Quantitative Genetics of Behavior (moderator: Bob Hitzemann)

4:30 - 5:00 Hugh Dingle "Quantitative genetics of life history strategies of insects."

5:30 - 6:00 Chris Boake "Sexual selection, speciation, and genetics."

6:00 - 6:30 Lisa Meffert "Nonadditive genetic effects in mating behavior."

July 8, 1999

8:30 - 10:30 Discussion (Moderators: Dingle and Sokolowski)

Questions:

1. Are QTL analyses of behavior and classical quantitative genetics compatible approaches?

2. How can classical quantitative genetics advance current trends in genomics?

10:30 - 11:00 Break

11:00 - 12:00 Discussion of Program Objectives (Moderators: Bob Hitzemann and Rob Page)

Questions:

 

1. What research resources are needed to promote molecular genetic studies of behavior in model and non-model organisms?

2. What levels of funding support are needed for gene-based behavioral research?

3. How can NSF promote a better exchange of information and technology between the disciplines of genomics, molecular neuroscience, and animal behavior?

4. Should NSF support a special initiative within the Biology Program directed at understanding the genetic architecture of variation in behavior? Would it attract sufficient proposals?

12:00 - 1:00 Lunch

1:00 - 5:00 Continue discussion