Towards precise genome editing, base editors have been developed by fusing catalytically compromised Cas9 with deaminase components, mediating C-to-T (cytosine base editors) or A-to-G (adenine base editors) transition. We developed a set of vectors consisting of a 5'-NG-3' PAM-recognising variant of SpCas9 with adenosine deaminases, TadA7.10 or TadA8e. Using a phenotype-based screen in Arabidopsis thaliana targeting multiple PDS3 intron splice sites, we achieved up to 81% somatic A-to-G editing in primary transformants at a splice acceptor site with NGG PAM, while 35% was achieved for the same target adenine with NGA PAM. Among tested vectors, pECNUS4 (Addgene #184887), carrying TadA8e, showed the highest ABE efficiency. With pECNUS4, we recreated a naturally occurring allele of DANGEROUS MIX3 (DM3) in two generations, transgene-free, for NGC PAM. We also simultaneously base-edited four redundant DM1/SSI4 homologs, encoding nucleotide-binding leucine-rich repeat (NLR) proteins, using a single gRNA with NGA PAM targeting the conserved yet functionally crucial P-loop motif of NLR proteins. We found fixation of A-to-G in three NLR genes for all three possible adenine sites within base-editing window 3-9, as the edited genes segregate in T2. Multigene targeting succeeded in rescuing the previously reported autoimmune phenotype in two generations. Mediating desired ABE on seven NLR genes simultaneously was successful as well; above 77% editing was achieved in six of the seven possible targets in a T1 plant, with the remaining having a moderately high (32%) editing. ABE application to specifically inactivate functional motifs is anticipated to expedite the discovery of novel roles for proteins.