Implications

Systems biology is a new way of working, which promises to increase the rate of progress across basic and applied biology.

The effect of plumage on poultry production and welfare

Feather coverage is an important factor in poultry production, influencing production, welfare and carcass processing. In broiler chickens the feathers comprise over 5% of the weight of the animal at slaughter (Cahaner et al., Reference Cahaner, Ajuh, Siegmund-Schultze, Azoulay, Druyan and Zárate2008), this feather weight being composed almost entirely of protein. Growth of a full plumage thus imposes a significant nutritional demand and necessitates removal after slaughter, producing millions of tonnes of feather waste per year.

The nutritional input required to produce a plumage, and its insulating effect, are generally acceptable or beneficial in developed countries with temperate climates; the conditions under which the modern highly productive lines of poultry have been bred. However, in hot conditions poultry suffer heat stress, which decreases production of eggs and meat, and can be lethal (Renaudeau et al., Reference Renaudeau, Collin, Yahav, de Basilio, Gourdine and Collier2012). The heat sensitivity of domestic chickens may appear unusual, given their tropical ancestry, but modern broilers gain weight much more rapidly than their antecedents, generating far more metabolic heat and having a bulkier body conformation that makes heat dissipation less efficient.

Studies of chicken lines with reduced plumage demonstrate that feather coverage makes a major contribution to overheating in hot conditions. The Naked neck trait, which is carried as a single co-dominant locus, exhibits bare neck skin and a reduction of feather coverage on the body of ∼20% in heterozygotes and 40% in homozygotes (Figure 1a). This trait has been in existence for some hundreds of years and now has a worldwide distribution, being particularly common in hot countries. Studies comparing Naked neck to fully feathered chickens report that above temperatures of about 30°C both egg and meat production, and feed efficiency, are greater than those of fully feathered chickens (Merat, Reference Merat1986; Yalcin et al., Reference Yalcin, Testik, Ozkan, Settar, Celen and Cahaner1997; Deeb and Cahaner, Reference Deeb and Cahaner2001; Singh et al., Reference Singh, Kumar and Singh2001).

Figure 1

A systems approach to understanding feather distribution in the chicken. (a) Feather development begins at 7 days’ incubation, with feather precursor cells being revealed by detection of CTNNB1 expression (purple colour). This process normally produces a plumage that covers most, though not all, of the skin. Naturally occurring variation at GDF7/BMP12 causes the Naked neck condition, whereas complete lack of feathering and foot scales is caused by lack of FGF20 function in the scaleless mutant. Determination of the molecular genetic basis of the Naked neck and scaleless traits identifies components of the embryonic feather patterning system. (b) Schematic of a simple network capable of producing periodic patterns, such as that of feather primordia in the embryonic skin. The Inhibitory process has a greater range of action than the Activatory, potentially through a greater rate of diffusion. Stimulation and suppression are indicated by + and − respectively. (c) Experimental suppression of GDF/BMP signalling in cultured chicken skin causes a pattern transition, which yields stripes, rather than spots. Computational simulation of an Activator–Inhibitor network reveals that this type of pattern change is expected when Inhibitor function is suppressed. The concordance between experimental and computer simulated outcomes supports the idea that GDF/BMP family members act as Inhibitors in a pattern-forming network with counterpart Activators, the latter potentially including FGF20. GDF = growth and differentiation factor; FGF20 = fibroblast growth factor 20.

A greater reduction of feather coverage than that of the Naked neck line is achieved by the scaleless mutation. This trait first appeared in an experimental flock maintained by the University of California and displays recessive inheritance (Abbott and Asmundson, Reference Abbott and Asmundson1957), though with variable penetrance depending on genetic background. On a suitable genetic background, this line lacks all body feathers and foot scales (Figure 1a). Studies comparing the entirely featherless scaleless chickens with their fully feathered siblings have found that weight gain achieved by the scaleless homozygotes at 6 weeks of age is greater than that of controls at high temperatures. In these conditions, the body temperature of featherless birds remains lower than that of their fully feathered siblings under the imposition of chronic heat stress, and in acute heat stress the featherless birds display far lower rates of mortality (Azoulay et al., Reference Azoulay, Druyan, Yadgary, Hadad and Cahaner2011). The effect of the complete lack of feathering caused by the scaleless mutation on body temperature and growth rate is greater than that of Naked neck in hot conditions (Cahaner et al., Reference Cahaner, Ajuh, Siegmund-Schultze, Azoulay, Druyan and Zárate2008). Thus, the optimal feather coverage varies with different environmental conditions, and benefits can be attained with plumage reductions of >60% of the normal feather coverage.

The genetic basis of variation in feather coverage: identification of system components

In order to identify specific genes that control the degree of feather coverage, we mapped the Naked neck and scaleless mutations. Naked neck was located in a region on chicken chromosome 3, initially identified by employing an experimental backcross and then by searching for a homozygous identical-by-decent region in outbred Naked neck individuals from distinct geographic locations. Within this region, we identified a large insertion of intergenic DNA of chromosome 1 origin into an intergenic region of chromosome 3. This insertion is unique to Naked neck individuals and is not present in wild type chickens. As we did not identify any mutations predicted to alter gene coding sequences, this suggested that the insertion acts to alter expression of nearby genes. We found that only one of the genes in the vicinity of the insertion, Growth and Differentiation Factor 7 (GDF7, also known as BMP12) is normally expressed in developing chicken skin, and that it displays increased expression in Naked neck embryos (Mou et al., Reference Mou, Pitel, Gourichon, Vignoles, Tzika, Tato, Yu, Burt, Bed'hom, Tixier-Boichard, Painter and Headon2011). This indicated that the Naked neck phenotype arises from increased expression of GDF7 due to alteration of the local gene regulatory landscape. The growth and differentiation factor proteins are intercellular signalling proteins known to be involved in a range of developmental and other processes. Interestingly, mutations in genes related to GDF7 have been implicated in production traits in a number of livestock species and organ systems, including loss-of-function mutations of GDF8 (also known as myostatin) as a cause of increased muscularity in cattle (Grobet et al., Reference Grobet, Martin, Poncelet, Pirottin, Brouwers, Riquet, Schoeberlein, Dunner, Menissier, Massabanda, Fries, Hanset and Georges1997) and sheep (Clop et al., Reference Clop, Marcq, Takeda, Pirottin, Tordoir, Bibe, Bouix, Caiment, Elsen, Eychenne, Larzul, Laville, Meish, Milenkovic, Tobin, Charlier and Georges2006) and mutation of GDF9 associated with an increased ovulation rate and frequency of twinning in sheep (Hanrahan et al., Reference Hanrahan, Gregan, Mulsant, Mullen, Davis, Powell and Galloway2004), suggesting that livestock can tolerate mutations in this class of genes in ways that sometimes confer beneficial production traits.

To identify the causative genetic lesion in the scaleless line, we mapped the trait in a flock segregating wild type and scaleless alleles using pooled blood samples from homozygous and heterozygous individuals. Single-nucleotide polymorphism genotyping of DNA from these two pools identified an association peak on chromosome 4, and we sequenced the open reading frame of all genes in this region to identify candidate mutations. We found a premature stop codon predicted to lead to loss of function in the gene encoding fibroblast growth factor 20 (FGF20) that is unique to the scaleless line (Wells et al., Reference Wells, Hadad, Ben-Avraham, Hillel, Cahaner and Headon2012). The FGF family was previously known to be important for feather development, making a molecule of this class a good candidate for the scaleless mutation (Song et al., Reference Song, Wang and Goetinck1996).

These molecular genetic studies identifying specific DNA alterations in Naked neck and scaleless lines can be informative for introgression of these traits into production lines, but they do not yet constitute a systems biology approach, focussing as they do on simple Mendelian traits rather than interactive processes. These findings do, though, identify components of the feather-forming system that operates in the skin and provide a means to study this process.

From genes to systems: predicting pattern change

In chickens, the number and distribution of feathers in the adult is defined in the skin before hatch, with periodically spaced clusters of embryonic skin cells assuming feather identity and then beginning outgrowth from the skin surface (Figure 1a). This process readily lends itself to study using a systems biology approach based on long-standing and mathematically well-supported theory about the general types of interactions that can produce such patterns. The pioneering work of Turing, Gierer and Meinhardt (Turing, Reference Turing1952; Gierer and Meinhardt, Reference Gierer and Meinhardt1972; Meinhardt and Gierer, Reference Meinhardt and Gierer2000) established the basic principles of systems that can produce periodic patterns, such as those seen in the arrangement of feathers in the skin. At its simplest, such a pattern-forming system consists of two processes, one acting locally and self-enhancing, this being coupled to a longer-range inhibitory influence. Although it is possible that each of these processes is embodied in the action of a single molecule, the theory does not stipulate this and allows for a number of physical entities to act together to yield the function of the ‘Activator’ and ‘Inhibitor’ that constitute the components of the system (Figure 1b). This Activation–Inhibition scheme, though an idealisation, provides a powerful framework for defining general functions of molecules involved in patterning and in searching for their interactions. In addition, decades of research into this type of system has led to the development of useful computational tools to study biological pattern formation.

On the basis of our genetic data, we considered GDF7 as a system component carrying out at least part of the Inhibitory process in Naked neck embryos. To infer aspects of the system of feather formation and explain its propensity to change, we investigated this patterning process using GDF7 protein as a reagent to interrogate how the system tends to change when the Activator–Inhibitor balance is altered. By treating cultured wild type skin with different concentrations of GDF7 protein, we found that it was possible to recreate the Naked neck phenotype ex vivo. The experimentally altered pattern was characterised by absence of neck feather formation and a dose-dependent reduction in extent of feathering on the body, with peripheral feathers being abolished at lower GDF7 doses than the central ones. This demonstrated that different regions of the body display different sensitivity to GDF7 action, with neck skin being particularly sensitive to its action in suppressing feather development.

We then performed computational simulations of the process of feather patterning, positing the action of interacting Activator and Inhibitor in producing the pattern. The running of simulations on which part of the field, representing the neck skin, is more sensitive to Inhibitor than the remainder of the field, reproduced the experimental results with respect to progressively decreased feathering on the body, but abrupt feather loss on the neck, in response to applied Inhibitor. Further, a prediction from simulation was that if Inhibitor function were reduced then stripes would form instead of spots, and that these stripes would be manifest with milder suppression of Inhibitor on the body than that required for their formation on the neck. We were able to use pharmacological inhibitors that suppress GDF-class signalling to test this prediction and found that pattern behaviour was indeed graded in the manner suggested by simulation, with spot enlargement, stripe formation (Figure 1c) and ultimately the formation of a single large feather precursor region over the entire skin (Mou et al., Reference Mou, Pitel, Gourichon, Vignoles, Tzika, Tato, Yu, Burt, Bed'hom, Tixier-Boichard, Painter and Headon2011). Thus, the specific pattern behaviours occurring upon experimental perturbation could be reproduced using a template of likely interactions, even without knowing all molecular components at this stage of investigation.

Putting the pieces together: defining molecular interactions

We next sought a molecular explanation for the increased sensitivity to GDF7 signals on the neck by performing expression array hybridisations comparing the transcriptomes of embryonic chicken body and neck skin. This comparison revealed that neck skin produces and experiences far more signalling through the retinoic acid pathway than does body skin. Retinoic acid is a vitamin A derivative, produced by the action of retinaldehyde dehydrogenase enzymes, which is essential for a range of developmental processes. We used the skin culture system to explore interactions between GDF7 and retinoic acid, finding that the latter sensitised tissue to the Inhibitory action of the former, allowing full suppression of body feather development by GDF7. Testing of regulatory interactions in culture also revealed suppression of FGF20 expression by the GDF-class of proteins, suggesting that FGF20 may act as the counterpart Activator to the GDF Inhibitors. However, further research is required to define the range of action and duration of these signals to explore the extent to which these molecules carry out the theoretical Activator and Inhibitor functions. Such quantitative data will be used to refine computer simulations of the feather-forming process, which may then be used to predict the properties of agents currently missing from our understanding of the system.

These basic biological studies are informing us of the types of changes in plumage patterns that chickens tend to undergo based on inherent developmental constraints. Even as we continue to identify components and interactions, the strong conceptual Activator–Inhibitor framework is useful in defining the aspects of the patterning process that we do not yet understand and is powerful in predicting the types of change possible. Part of our future work focusses on feeding this information back to whole-animal studies in order to understand the interaction of distinct genetic loci on feather distribution. This direction aims to inform the potential breeding of a wider range of feather coverage phenotypes, enabling a genetic approach to tailor the plumage to suit climactic or seasonal conditions, local housing and processing facilities and consumer preference.

Integrating genetic, genomic and network approaches to improving meat quality

A clear example of systems biology operating as a structure for data analysis was presented by Dr D.J. de Koning, Swedish University of Agricultural Sciences. His group used multiple interlocked technical and analytical approaches to understand the basis of variation in pH of broiler breast meat post-slaughter, an important indicator of meat quality. They first mapped the trait using a quantitative trait locus (QTL) approach, genotyping several hundred individuals to identify genomic regions associated with variations in meat pH. Then, a smaller number of homozygous individuals at a single QTL were selected for detailed investigation by transcriptional profiling of breast muscle. Mapping differences in gene expression between the genotypes by genome location identified a sharp peak that coincided with the QTL association peak, indicating that a cluster of genes showed coordinated change in expression and suggesting a regulatory mutation with effects on expression of many genes at this locus is responsible for the contribution of this genomic region to variation in meat pH.

Further, their gene network analysis implicated oxidative stress as the likely mode through which pH is altered by this QTL, indicating the metabolic route to low meat pH and highlighting not only genetic but also possible metabolic intervention to improve this condition. This work thus identified specific loci to allow selection on this post-slaughter trait in short term, but also identified genetic pathways and processes to consider in the longer term with broader applicability in broiler chickens. Notably, no overarching conceptual framework was adopted at the outset of this project, rather, the data were collected and analysed in unbiased fashion to infer the nature of the underlying process, distinguishing this work from that on feather patterning described above.

Systems approaches on a continental scale: improving lamb production in Australia

On another level of organisation, but using similar interrogative principles, the application of a systems approach to target efficiencies in lamb production on a national scale was presented by Dr Forbes Brien of the South Australian Research and Development Institute. The entire livestock production system under investigation involves a large element of human design, with detailed records providing a valuable data set for mining. Thus, and in contrast to the other two projects presented, at the outset of this work many of the components and interactions of the system under study were already known and the research questions undertaken related to quantitative analysis of the system to identify its weaknesses.

This project has the clearly defined objective of increasing sheep reproductive efficiency in Australia, specifically targeting the weaning of 1 million more lambs nationally by the year 2014. To progress towards this goal the team first determined the impact of life stage events from ram-ewe joining to lamb weaning in order to identify points where efficiency of production is low. This revealed a bottleneck at the level of poor neonatal (up to 72 h after birth) lamb survival, suggesting that focussing on increasing survival rates would make a significant impact on meeting the stated production target, as well as having clear animal welfare benefits.

In order to improve this specific aspect of lamb production, heritability estimates for lamb survival were determined. The direct heritability estimates for survival were very low, implying that genetic gains attained from selection on this characteristic would be attained slowly. Indeed, selection against non-viable lambs clearly occurs passively through their elimination from the flock. As there is likely a random component influencing which individuals are exposed to conditions sufficiently challenging to cause an early death, a major question is how to identify those lambs which are sufficiently robust to survive had they been exposed to the same environmental challenges as those which succumbed. This led to a consideration of means of indirect selection for survival, and measurement of a range of lamb traits found that crown–rump length, time to first bleat and rectal temperature show the strongest correlations with lamb survival. These traits have higher heritability than simple survival and are readily measured, thus serving as targets for indirect selection to improve lamb survival. In addition, maternal behaviour and temperament likely play a large role in lamb viability, so a major focus has shifted to understanding how to select for improved mothering ability. These modified selection criteria, together with provision of nutritional advice, is aiding producers towards meeting the increased lamb weaning target.

This programme represents a mature study in systems biology, based on the extent of knowledge of system structure at its outset. The explicit appearance of economic considerations and industry structure in the model contrasts with the previous, more biologically focussed, presentations. The potential complexity of understanding not only one but a number of interacting biological and economic systems was in this case greatly reduced by focussing on a clearly defined and quantitative goal.

Outlook

The unifying strand tying together research into these disparate problems lies in conceiving of systems in which interactions are as important as components, with the scale of organisation being of lesser importance. These approaches can contribute to the varied and often changing production goals required in modern agriculture. It is likely that application will come in a variety of forms; in the examples related to feather coverage and meat quality systems biology operates ‘behind the scenes’ solely as a research tool to identify individual interventions for application. In other cases, such as the programme focussed on improving lamb survival, the systems approach comes to the fore to analyse the production process itself, identifying sources of its fragility with the aim of building more robust production systems.

Investigation at a systems level in biology remains a young and developing area, though rapidly being adopted by a wide range of researchers with different backgrounds. Now that limitations on data availability and handling have been greatly reduced, a huge amount of knowledge will be generated by this trend in the research community, providing an opportunity for those who can find ways to harness the information surge for practical use.