Current predictions for food security in the year 2050 look pretty dire to be honest. If the global population reaches the forecasted ~9 billion, then our overall food production will somehow have to increase by 70% between now and then. And that doesn’t even begin to factor in potential crop production for biofuels. (There was me thinking things could only get better after 2016.)
However, it’s not all doom and gloom. Existing technologies and resources offer considerable potential to increase global yield and expand onto unused land. Advances in genetic modification (GM) and breeding techniques, combined with modern fertilisers and pesticides offer ways to tackle the major enemies of crop yield – pathogens, weeds, poor nutrition and environmental conditions. While crop yield is often determined by a complicated interplay of a number of these factors, plant disease stands out as a more obvious issue to take on, with some estimates attributing upwards of 10% global yield reduction to pathogens alone.
Efforts in recent years have focussed on using genetic engineering to introduce resistance (R) genes from wild ancestor species and more distantly related species into modern day crops. R genes function at the ‘second level’ of plant immunity after pathogen-associated molecular pattern (PAMP)-triggered immunity fails. R genes work through producing proteins that recognise specific pathogenicity effectors – proteins secreted by the pathogen that aid infection – in a gene-for-gene relationship.
The issue with R gene resistance is that, more often than not, it’s relatively short-lived. Rapid pathogen evolution coupled with selection pressure from R genes means effectors quickly evolve to circumvent this level of disease resistance. The research community has come to meet this problem head-on by developing new techniques that can give further insight into plant-pathogen interactions and ultimately new resistance strategies, as well as exploring the great diversity of R genes that we still know very little about. I wish I had the time and space – and frankly the brains – to examine all of the interesting strategies being utilised in plant pathology at the moment, but for now I thought I’d focus on two approaches that I find particularly interesting: data mining and image-based phenotyping.
Data mining exploits the large amount of genomic and transcriptomic data that is now available thanks to cheap and fast sequencing technologies. While many of these datasets remain only partially annotated, it is possible to ‘mine’ for resistance genes by comparing genomic and transcriptomic data between closely related species or by looking for phenotype/genotype associations. Data mining has already been successfully used to identify novel R genes and other potential resistance-associated genes in a number of important crops, such as rice tomato, and has also been applied in studies using machine-learning to predict crop pest trends.
Analysis of pathogen genomes and secretomes can likewise give useful data. There is such a great diversity of pathogen species and strains, deploying various infection mechanisms, that pinpointing important pathogenesis genes through mutant screens and other means would be a somewhat daunting task. The in silico approach allows multiple datasets to be compared and analysed at once, which can help single out promising genes for investigation. Once new effectors and small RNAs (sRNAs) with potential effects on host plant immunity are found, they can be added to the list of possible molecular targets to be utilised in resistance strategies and – to come full circle – can be used to inform future genome searches.
However, data mining isn’t perfect and applying the information gleaned from this technique is undoubtedly far more complicated than I just made it sound. Computational limitations are still a major issue, especially when dealing with multiple large datasets or trying to simulate biological processes. Moreover, data mining, as suggested by the name, is highly reliant on readily available and reliable data on the organisms of interest. Although there are growing number of databases online dedicated to crop disease and resistance information, varying quality and availability of data is still hindering research on certain diseases. (As a molecular biologist, I think it’s also important to note that at the end of the day, in vivo functional characterisation of genes is still always required.)
Another significant problem delaying the application of data mining information is a lack of knowledge about genotype-phenotype relationships in plants. This is largely due to phenomics research not being able to keep up with the rapid advances that genomics has made in recent decades. Traditional phenotyping methods rely heavily on being able to identify obvious disease symptoms such as lesions and wilting, but this can be skewed by assessor bias and fails to capture subtle phenotypes such as pathogen movement and growth within the plant. Image-based phenotyping allows these less evident plant-pathogen interactions to be visualised, as well as capturing information naked to the human eye through hyperspectral and thermal imaging. High-throughput automated imaging systems can generate and analyse large volumes of data, making it highly applicable to field settings, for instance through the use of drones or Smartphones.
A recent example that nicely highlights the use of quantifying spatial and temporal patterns of disease symptoms through image-based methods looked at cassava bacterial blight which is caused by Xanthomonas axonopodis (Xam):
Quantification of disease symptom development and bacterial spread was achieved through using a Raspberry Pi and camera board alongside looking at bioluminescent Xam strains. Once the images were generated, they were analysed using a semi-automated system, such as macro scripts on ImageJ. By using Xam strains mutated in three different Type III effectors (T3Es – the previously characterised AvrBs2 and XopX, and a putative effector selected from genomic data, XopK), they were able to look at the relative contribution of individual effectors to virulence and gain insight into their function. For instance, they showed that xopK knockout mutants induced disease symptoms more rapidly than wild type, seeming to contradict it’s predicted T3E identity. However, xopK mutants also exhibited reduced pathogen spread within the host plant, indicating that it may in fact play some role in virulence.
Their investigation of XopK not only neatly demonstrates the potential of image-based techniques to capture ‘additional dimensions’ to phenotypes, but is also a great example of how data mining and image-based phenotyping can be used in conjunction to produce powerful results. As stand-alone techniques, I’m afraid they’re probably not the all-encompassing solution to global food security (that we can only hope for), but it’s encouraging that scientists are adopting innovative approaches to upgrade and bolster agricultural research.
Feeding the world in 2050 will be no small feat, and will require changes to the way we farm food, not just the type of food we farm. While researchers may be making leaps and bounds towards producing sustainable crops in the lab and in field trials, there is still a lot of necessary red tape to get through before any real-world impact will be seen. We need governments to really start prioritising long-term sustainability policies that will allow this research to see the light of day and we, as the ones with the real power, need to really push forward with public engagement in order to remove the stigma that surrounds GM and other technologies, which we will undoubtedly need in the future. Oh, and if we could also stop putting climate change-deniers in positions of power, that would be great.
Pumplin, N. and O. Voinnet (2013). “RNA silencing suppression by plant pathogens: defence, counter-defence and counter-counter-defence.” Nat Rev Micro 11(11): 745-760.
Mahesh, H. B., et al. (2016). “Indica rice genome assembly, annotation and mining of blast disease resistance genes.” Bmc Genomics 17(1): 242.
Mutka, A. M, et al. (2016) “Quantitative, image-based phenotyping methods provide insight into spatial and temporal dimensions of plant disease.” Plant Physiol. 172(2): 650-660
Dangl, J. L, et al. (2013) “Pivoting the Plant Immune System from Dissection to Deployment.” Science. 6147(341): 746-751
(For anyone interested in finding out a bit more about R genes, this figure is a great starting point: http://www.nature.com/nrmicro/journal/v11/n11/box/nrmicro3120_BX2.html )