Anthracocystis panici-leucophaei e Bipolaris yamadae como agentes de biocontrole para o capim-amargoso (Digitaria insularis)

Abstract

Weeds are one of the main problems in world agriculture because they compete with crops for nutrients, water, light and space, and can also host pests and diseases and exert allelopathic effects on cultivated plants. One of the most harmful weed species in Brazilian agriculture is bitter grass (Digitaria insularis) due to the emergence of biotypes resistant to glyphosate and ACCase-inhibiting herbicides. Biological control then becomes an important tool for the integrated management of this weed. Anthracocystis panici-leucophaei has shown to be a promising fungus for the management of this weed, as it inhibits the growth of bittergrass plants and significantly reduces its seed production. Thus, the objective of this work was to evaluate the aggressiveness of 16 isolates of A. panici-leucophaei as potential biocontrol agents for bittergrass, to evaluate the host range of this species, as well as to develop, using artificial intelligence, a model for identifying the inhibition of growth in D. insularis plants caused by systemic infection of the fungus, and to identify, through morphological and phylogenetic analysis, a fungal isolate (COAD 3424), found causing leaf spots and blight in D. insularis leaves . Plants with smut symptoms were collected in the states of Minas Gerais and Goiás. To evaluate the aggressiveness of different isolates of A. panici-leucophaei, pure cultures of the isolates were obtained in the laboratory for inoculation in newly emerged bittergrass plants. Isolate 46 of this fungus was the most aggressive among the tested isolates. To evaluate the specificity, the following species were tested: Saccharum officinarum, Oryza sativa, Triticum spp, Sorghum bicolor, Zea mays, Pennisetum glaucum, Eleusine indica, Urochloa hybrida, Urochloa brizantha, Urochloa ruziziensis, Panicum maximum and Cynodon spp. Anthracocystis panici-leucophaei was specific to D. insularis. For the development of a model to identify the effect of the systemic infection caused by this fungus, the artificial intelligence algorithms Random Forest, Support Vector Machine and Multilayer Perceptrom were used. It was possible to create a model to predict the growth inhibition of bittergrass plants inoculated with spores of A. panici-leucophaei and the algorithm that obtained the highest accuracy was Random Forest. The fungal isolate (COAD 3424) was identified as Bipolaris yamadae, being the first report of this fungus in D. insularis in the world.