Progress 06/15/20 to 12/31/23
Outputs
Target Audience:For the duration of the project, I was able to reach several fungal rust researchers, plant pathologists, plant biologists, plant breeders, maize researchers, and students studying biology, machine learning, and computer vision applications for plant research. Changes/Problems:Originally, we aimed to identify maize factors that interacted with our predicted P. sorghi effectors through yeast-two hybrid screening. Identified maize factors were to be silenced via VIGS with FoMV. Due to the inefficiency in the first library, as well as the strong autoactivation of our most promising candidate, 930g11, in our second library, we instead opted to silence maize factors identified as important to this pathosystem from another time course experiment. We were able to create twelve VIGS constructs of these maize factors and had enough time to start the first VIGS experiment. We did not have enough time to process the results, but RNA samples and image scans were taken of each leaf in the experiment to assess silencing ability, effect on other maize leaves, and phenotype of the disease symptoms on each plant with our developed machine learning pipeline. What opportunities for training and professional development has the project provided?Throughout the course of this project, I was able to attend several seminars and conferences, many of which offered specific professional development workshops that I attended. At these seminars, I was able to present my own research, both in oral and poster presentations, as well as learn from others' research. I learned many new protocols, lab techniques, and skills from other lab members and collaborators. Additionally, through the course of my research itself, I was able to advance my understanding of bioinformatics and machine learningand apply it directly to my own research. This project also enabled me to directly mentor two undergraduate research assistants, working with them one-on-one to teach them about lab techniques, protocols, and experimental design. I was able to work with several others in the lab as both a mentor and mentee. This project also enabled me to earn my PhD in Plant Biology. How have the results been disseminated to communities of interest?Results from this project's research have been shared through several seminars and conferences in both oral and poster presentations throughout the project, which included plant biology graduate students and rust, maize, and genetics researchers. Although a long-read genome assembly for the Puccinia sorghi IA16 isolate has not been published yet, the draft assembly and list of putative proteins have been shared with several fellow rust researchers for use in their own research. The formal publication of this resource is in progress. A peer-reviewed journal article detailing the generation and use of the machine learning model developed for Objective 3 is now available. Additionally, all data used in the publication is maintained in a public repository, which includes the raw image files, the annotated regions for all pustules, and details about the dataset. The machine learning model from the publication and details regarding its use are available in a public GitHub repository. Both repositories are linked to the publication. My Ph.D. dissertation detailing much of the work done for this project is also publicly available. I also have a public GitHub repository detailing the assembly and usage of the timelapse phenotyping boxes written in accessible language. Protocols developed throughout the course of this project have been shared with several others, both in written form and in one-on-one training. What do you plan to do during the next reporting period to accomplish the goals?
Nothing Reported
Impacts
What was accomplished under these goals?
IMPACT: The overall goals of this project were to understand the Puccinia sorghi-maize pathosystem furtherand develop better methods, tools, and resources with which to do this. Fungal rust pathogens, including P. sorghi, pose a significant threat to crops worldwide, and a better understanding of the complex interactions between pathogen and host may lead to more substantial and more robust control strategies, both specific to P. sorghi and for other rust species. Adoption of the resources and knowledge gained from this project will enable researchers to conduct more quantitative tests, make better decisions on the targets they choose to study, and enable comparative studies for various rust isolates, which differ in their interactions with various maize varieties. Objective 1: We acquired Nanopore long-read sequencing data for the P. sorghi isolate IA16 and HiC sequencing for isolates IA16 and IN2. Utilizing the Nanopore reads, we generated a long-read genome assembly for the IA16 isolate. For the last version of this genome assembly, we first updated the basecalls on all long-read sequencing with the most current basecaller, which improved contiguity without requiring additional sequencing efforts. The assembled genome from these updated base calls has 620 scaffolds composed of 1,159 contigs, totaling 166 Mb in sequence, of which 76% is predicted to be repeat elements. The use of long-read sequencing, enabling a more contiguous genome and better resolution of repeat regions, gave insight into the actual genome size and repeat element sequence of the P. sorghi genome. Utilizing RNAseq data specific to the IA16 isolates, we were able to annotate 18,520 predicted genes using a de novo approach. The use of isolate-specific RNAseq data is an improvement over currently available resources, which relied on predicted genes from other species as evidence for gene annotation. Of the predicted genes, 1,033 are predicted to be candidate-secreted effector proteins (CSEPs), many of which are unique to this assembly. This provides additional effector candidates for screening and characterization purposes. Due to reduced output from the Nanopore flow cells during sequencing, time-intensive spore collection to acquire enough sample to extract high-quality genomic DNA from, and the high cost of additional sequencing, we were unable to fully phase the genome assembly. However, this genome, as it stands, is an improvement over previous resources and is already being utilized by other rust researchers. Furthermore, this assembly shows that a relatively complete and contiguous genome can be assembled with limited long-read sequencing data that can be further scaffolded with HiC sequencing data from an organism that is traditionally difficult to extract genomic DNA from. Objective 2: Cluster 112 is a group of putative rust-specific effector proteins, of which at least one member from Phytophthora pachirhizi, PpEC23, has been shown to affect plant immunity. We identified eleven CSEP candidates related to PpEC23 in P. sorghiand were able to amplify eight of those candidates from cDNA. These eight candidates were used in several immune suppression assays in Nicotiana benthamiana leaves. In the first assay, the eight candidates were transformed into Pseudomonas syringae pav. Tomato DC3000, which usually triggers a strong hypersensitive response (HR) in N. benthamiana leaves. Only CSEP 930g11 showed significant suppression of HR, with two additional CSEPs showing no significant impact and the remaining five showing increased HR symptoms. Since 930g11 was the only CSEP showing significant suppression of plant immune response in the previous assay, we acquired transgenic N. benthamiana lines constitutively expressing the CSEP 930g11. These lines were selfed twice, and azygous plants from four of the original lines were saved for use as negative controls. We also made simple phenotyping boxes utilizing Raspberry Pis to take time-lapse images of experiments with these transgenic plants. The boxes themselves are inexpensive to make, and extensive details about their setup and use are available in a public GitHub repository we created. Immune suppression assays using these 930g11 overexpressing lines found that although not statistically significant, the transgenic leaves consistently showed a lower HR when compared to the azygous controls for all independent transgenic lines. Reactive oxygen species (ROS) burst assays were also conducted on these transgenic 930g11-expressing lines to examine one aspect of pathogen-triggered immunity (PTI). In this assay, 930g11 does not appear able to suppress flagellin 22-triggered ROS burst. Taken together, it seems that 930g11 has a small suppressive effect on effector-triggered immunity. The use of the phenotyping boxes allowed us to generate timelapse image data that were then used to have more quantitative results for these experiments. Additionally, detailed instructions regarding their setup and their low cost of assembly means other researchers can easily implement similar setups in their own research. Additionally, a new yeast two-hybrid (Y2H) library was generated to discover the interacting proteins of the P. sorghi-maize pathosystem. We created bait yeast strains for five of the Cluster 112 CSEPs, including 930g11. Unfortunately, 930g11 was a strong autoactivator, and we were unable to mate it to our prey library. We mated two other candidates to the library with inconclusive results. Regardless, the Y2H prey library will be useful for future experiments, both with the remaining P. sorghi Cluster 112 CSEPs and with new targets identified from the genome assembly generated in Objective 1. Objective 3: Since common rust symptoms of maize leaves are small and numerous, we wanted to develop a quantitative way to measure disease without manual counting or biological assays. A machine learning pipeline was developed utilizing a U-Net neural network. This work assessed the effects that specific training images and different amounts of training data have on a particular model's performance. This work also resulted in a final model trained with all available data to use for the phenotyping of common rust symptoms on maize leaves in future experiments. Overall, the majority of models generated were able to replicate ground truth results. The final model that will be used moving forward gives results that trend similar to ground truth results. This model and all data used to train it are publicly available, and their use is expected to make quantitatively phenotyping rust disease symptoms easier and faster. A peer-reviewed publication detailing the model and its results is also available. The details in the publication may help inform researchers looking to apply a similar concept to other pathosystems. The entry point for utilizing our generated model is also very low, as it requires minimal expertise and funding to implement, meaning we expect it to be useful for a wide variety of researchers. Although we were unable to identify maize-interacting factors in the previous objective, we were able to identify maize factors thought to be important to the P. sorghi-maize pathosystem through RNA and protein network analysis data shared with us. From these data, we chose twelve candidate maize genes and generated VIGS constructs for each. Preliminary VIGS experiments were conducted with all 12 constructs in the maize lines Rp1-D, which is resistant to P. sorghi isolate IA16, and H95, which is susceptible to IA16. RNA samples and images of leaves were taken of each plant for further analysis. This experiment laid out a pipeline for similar experiments. As more factors are identified through effector characterization and Y2H screening, they can be moved through this pipeline for streamlined analysis and phenotyped quickly and quantitatively with our developed machine-learning model.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2024
Citation:
Holan, K. L., White, C. H., and Whitham, S. A. (2024) Application of a U-Net Neural Network to the Puccinia sorghi Pathosystem. Phytopathology.
- Type:
Theses/Dissertations
Status:
Published
Year Published:
2023
Citation:
Holan, K. L. (2023). Genomic and phenomic approaches for studying Puccinia sorghi-maize interactions (Publication No. 2859464478) [Doctoral dissertation, Iowa State University]. ProQuest Dissertations & Theses Global.
|
Progress 06/15/22 to 06/14/23
Outputs
Target Audience:During this reporting period, I was able to reach students studying computer-vision-based methods for plant phenotyping, particularly for plant disease phenotyping. I was also able to reach plant disease researchers, including fungal rust researchers, plant immune researchers, and general plant biology researchers. Changes/Problems:Originally, we aimed to identify maize factors that interacted with our predicted P. sorghi effectors through yeast-two hybrid screening. Identified maize factors were to be silenced via VIGS with FoMV. Due to the inefficiency in the first library, as well as the strong autoactivation of our most promising candidate, 930g11, in our second library, we instead opted to silence maize factors identified as important to this pathosystem from another time course experiment. The end result of this change is expected to be the same, where maize factors important to pathogenicity may be identified. What opportunities for training and professional development has the project provided?I attended several seminars and conferences, both to present my own research as well as learn about others' research. I was able to learn new protocols, lab techinques, and skills from other lab members and collaborators. Additionally, I was able to advance my understanding of bioinformatics, as well as apply it directly to my research. How have the results been disseminated to communities of interest?Several seminars and poster presentations have been used to disseminate results from this research. Additionally, two GitHub repositories have been made available, one relating to the immune assay phenotyping boxes and the other relating to the machine learning pipeline for common rust disease quantification. The images and annotations used in the machine learning pipeline are currently being made publically available for other rust researchers to use as well. What do you plan to do during the next reporting period to accomplish the goals?1. Currently, the IA16 genome assembly has been assembled and annotated. We aim to increase the quality of those annoations with protein data currently available to us. The final assembly is expected to be uploaded to GenBank and details regarding the assembly published soon after. 2. Yeast-two hybrid mating of effector candidates will be conducted. 3. Data from the VIGS experiments will be processed to identify any phenotypic effects of silencing of the maize genes. Additionally, the full U-Net machine learning pipeline will be reported on.
Impacts
What was accomplished under these goals?
IMPACT: The main problem addressed by this project is to further understand the Puccinia sorghi-maize pathosystem. Fungal rust pathogens, including P. sorghi, pose a significant threat to crops worldwide, and a better understanding of the complex interactions between pathogen and host may lead to stronger and more robust control strategies, both specific for P. sorghi and, more largely, for other rust species. In addition to a better understanding of this pathosystem, this project aims to develop both datasets and methods aimed at helping other rust researchers with their own projects and analysis. Specific target proteins in both the pathogen and the maize host are expected to be of importance to maize breeders and disease and plant immune response researchers. Protocols for machine learning-based and image-based phenotyping for greenhouse-scale research are expected to be of use to both non-experts and those directly involved with developing plant phenotyping methods. Objective 1. Additional long-read sequencing was generated for the P. sorghi isolate IA16 assembly, and a new draft genome was assembled. Repeat regions and genes were predicted and annotated within the assembly. The draft genome has 902 scaffolds and ~16,000 gene models, of which 742 are predicted to be secreted effector proteins. Objective 2. Additional immune suppression assays were conducted on the candidate-secreted effector protein 930g11-expressing transgenic Nicotiana benthamiana lines. There was a consistent suppression of immunity for four lines of transgenic N. benthamiana plants when compared to their azygous counterparts, consistent with previous results for this candidate effector. Reactive oxygen species (ROS) burst assays were also conducted on these transgenic lines to look at one aspect of pathogen-triggered immunity (PTI). In this assay, 930g11 does not appear able to suppress flagellin 22-triggered ROS burst. Additionally, a new yeast two-hybrid library was generated to discover the interacting proteins of the P. sorghi-maize pathosystem. Objective 3. A new machine learning pipeline was developed utilizing a U-Net neural network. This work assessed the effects that specific training images and different amounts of training data have on a particular model's performance. This work also resulted in a final model trained with all available data to use for the phenotyping of common rust symptoms on maize leaves in future experiments. Overall, the majority of models were able to replicate ground truth results. The final model that will be used moving forward gives results that trend similar to ground truth results. Additionally, other researchers uncovered maize factors that seem to be very important to the P. sorghi-maize pathosystem. A short list of maize factors was chosen, virus-induced gene silencing (VIGS) constructs using a foxtail mosaic virus (FoMV) vector was created, and VIGS experiments were conducted. Briefly, maize seedlings were inoculated with FoMV vectors containing sequences from the identified maize factors, inoculated with P. sorghi after VIGS was suspected of having started, and those leaves were scanned to run through the developed machine learning pipeline.
Publications
|
Progress 06/15/21 to 06/14/22
Outputs
Target Audience:This reporting period, I was able to reach students studying biology, machine learning, and computer vision applications for plant research. Additionally, I was able to reach plant disease researchers (including rust researchers), plant pathologists, machine learning application groups, and plant virology groups. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?I attended several seminars related to my research both online and within my university. In addition, I was able to work one-on-one with an undergraduate student to train them inlab and research skills. I was also able to learn new protocols and lab techniques from one-on-one training with others in the lab, as well. How have the results been disseminated to communities of interest?Results have been disseminated through two conference papers published for machine learning conferences (Artificial Intelligence and Soft Computing 2021 & 2021 International Joint Conference on Neural Networks). A poster was shared at the 2021 Plant Biology Meeting. Data and materials related to theP. sorghigenome assembly was shared via direct communication with various rust researchers. What do you plan to do during the next reporting period to accomplish the goals?1. Previously, long-read Nanopore sequencing was generated for the P. sorghi isolate IA16 and short read Hi-C sequencing data was generated for two isolates, IA16 and IN2. In order to have a largely phased genome as stated in the goals and to improve completeness and base quality, additional DNA extraction and long-read sequencing is in progress and is expected to be completed in the next few weeks. Additional long-read sequences for IA16 will be produced, alongside IN2 long-reads, enabling complete, phased, and highly contiguous assembliesfor two distinct P. sorghi isolates. Previous Nanopore flow cells had very low outputs due to complexities inherent in rust species. Newer DNA extraction protocols have been made available since the previous sequencing efforts, which shouldincrease output and quality of the reads. Once the genomes are assembled, they will be uploaded to NCBI and informationregarding their assembly will be published. 2. The previous attempt to construct a yeast two-hybrid library to discover the interacting proteins of the P. sorghi-maize pathosystem had very low efficiency and therefore was concluded to be significantly incomplete. Thus, a new library construction attempt will be made to find potential targets for interacting proteins. Given the inconclusive nature of previous immune suppression assays using the transgenic N. benthamiana lines, additional experiments will be conducted using another two lines, in addition to other experiments, including ROS burst assays to corroborate the immune suppression phenotype seen in previous results. 3. The newer machine learning pipeline results will be compared to the previous machine learning pipeline, and these results will be published alongside the finished model(s) for use by others. In addition to the models, the images used to train the models will be shared, allowing others to use these images to augment their own datasets or as a jumping off point forfuture phenotyping research/model training. Once targets from the yeast two-hybrid library have been identified, virus-induced gene silencing experiments can commence using the previously published injection method and Foxtail mosaic virus.
Impacts
What was accomplished under these goals?
IMPACT: The main problemaddressed by this project is to further understand the Puccinia sorghi-maize pathosystem. Fungal rust pathogens, includingP.sorghi, pose a significant threat to crops worldwide, and a better understanding of the complex interactions between pathogen and host may lead to stronger and more robust control strategies, both specific for P.sorghi and, more largely, for other rust species. In addition to a better understanding of this pathosystem, this project aims to develop both datasets and methods aimed at helping other rust researchers with their own projects and analysis, some of whom are currently using datasets developed during the course of this project. Specific target proteins in both the pathogen and the maize host are expected to be of importance to maize breeders and disease and plant immune response researchers. Protocols for machine learning-based and image-based phenotyping are expected to be of use to both non-experts and those directly involved with developing plant phenotyping methods, particularly for small groups on agreenhouse scale. Objective 1. A draft genome for P. sorghi has been completed, including partial phasing of the two nuclei utilizing both Nanopore long-read sequences and short-read HiC data. In its current state, the draft genome has ~1,000 contigs, which is substantially less than the current draft genome. Using BUSCO scores as a metric, the assembly seems to be relatively complete. Objective 2. Candidate secreted effector protein (CSEP) 930 transgenic Nicotianabenthamiana lines (which transgenically express the previously identified immune suppression candidate CSEP 930) underwent two rounds of inbreeding. In addition, azygous lines from 4 original lines were generated to act as controls. Several immune suppression assays using Pseudomonas syringae pv. tomato (Pst)DC3000 were conducted on these transgenic lines. Simple phenotyping boxes were developed that use Raspberry Pis to take timelaspe images of these leaf assays. These phenotypes were scored on their apparent immune response. Using two transgenic CSEP 930 lines and their respective azygous counterparts, 3 experimental replicates were conducted, leading to mixed results where one line showed apparent immune suppression and another showed apparent immune activation. Given the highly inconsistent nature of immune assays, these results are not wholly unsurprising and additional experiments, as well as testing on the additional two transgenic lines with azygous counterparts, are expected to better understand thisresponse. In addition,ROS (reactive oxygen burst) burst experiments on the transgenic lines have begun in order to provide an additional approach to understanding the effect CSEP 930 has on plant immunity. Objective 3. Previous machine learning projectshave been reported on and disseminated as conference papers. Using the model from the collaboration, data from two biologicalexperiments were conducted to then test its accuracy and useability for rust researchers. The first experiment utilized two P.sorghi isolates and a particular maize resistant line. One isolate is able to successfully infect this particular maize line while the other is not.. The model was challenged with leaf scans from 3 replicates. Although quantitative counting of individual pustules was inconsistent, the model was able to show the same statistical significances between groups as manual counts. The second experiment used varying levels of fungicide to induce arange in the density of pustules on leaves. The model was again challenged with leaf scans from 3 replicates, and like the previous experiment, the quantitative results of pustules counts wereinconsistent, but the model showed the same statistical significances between groups as the manual pustule counts. Encouraged by these results but hoping for more accurate quantitative results in regards to true pustule counts, a collaboration was started with another machine learning expert to develop an additional model using a different type of neural network. So far, results seem equal to or better than previous results.
Publications
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2021
Citation:
Pillay, N., Gerber, M., Holan, K. L., Whitham, S.A., Berger, D.K. (2021). Quantifying the Severity of Common Rust in Maize Using Mask R-CNN. In: Rutkowski, L., Scherer, R., Korytkowski, M., Pedrycz, W., Tadeusiewicz, R., Zurada, J.M. (eds) Artificial Intelligence and Soft Computing. ICAISC 2021. Lecture Notes in Computer Science(), vol 12854. Springer, Cham. https://doi.org/10.1007/978-3-030-87986-0_18
- Type:
Conference Papers and Presentations
Status:
Published
Year Published:
2021
Citation:
M. Gerber, N. Pillay, K. Holan, S. A. Whitham and D. K. Berger, "Automated Hyper-Parameter Tuning of a Mask R-CNN for Quantifying Common Rust Severity in Maize," 2021 International Joint Conference on Neural Networks (IJCNN), 2021, pp. 1-7, doi: 10.1109/IJCNN52387.2021.9534417.
|
Progress 06/15/20 to 06/14/21
Outputs
Target Audience:This reporting period, I was able to reach several rust researchers, plant pathologists, plant biologists, plant breeders, and maize researchers. I was also able to reach machine learning groups, plant phenotyping researchers, and those interested in long-read sequencing. Changes/Problems:
Nothing Reported
What opportunities for training and professional development has the project provided?I attended several smaller seminars and symposia during the last year, including Interdepartmental Plant Biology, Plant Pathology and Molecular Biology, and Crop Bioengineering Center seminars, the Interdepartmental Genetics and Genomics symposium, and Predictive Plant Phenomics learning communities and discussion. I have also been able to get one-on-one training from others in the lab for new techniques. How have the results been disseminated to communities of interest?Results have been shared through posters at the 2020 Maize Genetics Meeting and the 2020 Plant Biology conference. Additionally, a protocol was published in the Journal of Visualized Experiments demonstrating the agroinjection protocol for inoculating maize seedlings with virus-induce gene silencing constructs. What do you plan to do during the next reporting period to accomplish the goals? The phased, highly contiguous P. sorghi genome is expected to be finished during the next reporting period. This will be disseminated to the research community by uploading the data to NCBI as well as publishing a paper in regards to this assembly. Immune suppression assays will continue to be conducted on the CSEP 930 transgenic N. benthamiana lines. Development on the new timelapse phenotyping boxes will also continue. 930, along with at least two additional CSEPs that showed promise in the immune assay experiments (1483 and 2734) will be used as bait for the maize-P. sorghi yeast two hybrid library to identify potential interacting proteins. Upon the identification of interacting proteins, BiFC and Co-IP pull-down experiments to confirm these interactions will begin. Development of the machine learning pipeline will continue. A paper disseminating this pipeline and its usefulness on answering biological questions will be published. Once interacting proteins from Objective 2 are identified, VIGS development can begin on these proteins in maize.
Impacts
What was accomplished under these goals?
IMPACT: The overall goals of this project are to better understand the Puccinia sorghi-maize pathosystem. Fungal rust pathogens likeP. sorghiare very complex both in their lifecycles and in their interactions with plant hosts. By looking at a promising family of effector proteins ofP. sorghi, I hope to better elucidate the interactions and functions these proteins have in regards to their growth on maize. This work will generate data and protocols that will help other rust researchers with their own work. Additionally, understanding effector protein targets may help maize breeding programsgenerate more resistant varieties. So far, this project has generated data that can be used by other rust researchers. Additionally, at least one effector proteinwith asignificant impact on plant immunity has been identified, and this protein, along with a few other promising candidates,is moving forward for further study. Methods to better phenotype and gather quantitative data about disease progression in this pathosystem are also being developed, which will have applications both in plant-pathogen interaction research and plant breeding research. Objective 1.Develop a highly contiguous and completeP. sorghigenome sequence by long-read sequencing. Nanopore sequencing has been completed for the CR-IA16 P. sorghi isolate. In addition to long-read sequencing, HiC libraries were sequenced for the IA16 isolate and for an additional isolate, IN2, which varies in its differential profile on maize lines with varying resistance when compared to IA16 and therefore likely varies in its effector repertoire as well. A preliminary genome sequence has been assembled from these data with the Flye assembler. Objective 2.Characterize cluster 112 CSEPs to identify functions in suppressing host immunity and host target proteins. Several initial characterization experiments of the Cluster 112 CSEPs have been completed. Immune suppression assays in Nicotiana benthamiana leaves, where immune response to a CSEP is compared to that of Pst DC3000 were performed. From these experiments, a particular CSEP, dubbed 930, was found to significantly decrease the hypersensitive response in N. benthamiana. To further test immune response, and for experiments in protein-protein interactions, transgenic N. benthamiana lines constitutively expressing the CSEP 930 were generated. Preliminary experiments looking at immune response to Pst DC3000 have begun in these lines utilizing a timelapse phenotyping box. Additionally, construction of a maize-P. sorghi yeast two hybrid library has been started. Objective 3.Silence expression of host proteins that interact with cluster 112 CSEPs using virus-induced gene silencing and phenotype the resulting rust disease development with image-based computer vision techniques. Further development and optimization of the agroinjection method of introducing virus-induced gene silencing (VIGS) constructs to maize seedlings has led to a publication on this technique. In collaboration with the aforementioned group from the University of Pretoria, there has been development on a machine learning pipeline to more accurately phenotype P. sorghi on maize leaves.
Publications
- Type:
Journal Articles
Status:
Published
Year Published:
2021
Citation:
Beernink, B. M.*, Holan, K. L.*, Lappe, R. R., Whitham, S. A. Direct Agroinoculation of Maize Seedlings by Injection with Recombinant Foxtail Mosaic Virus and Sugarcane Mosaic Virus Infectious Clones. J. Vis. Exp. (168), e62277, doi:10.3791/62277 (2021).
|
|