© 2012 Nature America, Inc. All rights reserved.

594 volume 30 number 7 July 2012 NATURE BIoTECHNoloGy poly (NG)(specificity of NG is for thymine) and other RVDs did not activate any of the reporter constructs. We modified the naturally most variable non-RVD amino acids (positions 4 and 24) of NI, NG and NN repeats, but this did not restore activity of weak RVDs (Supplementary Fig. 2). This demonstrates that TALEs with certain RVD compositions are not functional, a finding that is contradictory to the simple TALE code reported so far2, 5, 6.

Possibly RVDs have different efficiencies and having many repeats with weak RVDs impedes TALE function. Therefore, we constructed ArtTALs, with an ‘anchor’consisting of the first six repeats of the natural TALE Hax3, which contain several RVDs HD, which are efficient according to our ArtTALpolyHD data (Fig. 1a). These six repeats were followed by six repeats with identical RVDs to yield a total of 11.5 repeats (Supplementary Sequences). Reporter constructs for in planta GUS assays contained the first half of the Hax3 box and six A, C, G or T, respectively. We used this artificial setup to test known and new RVDs. In total, we analyzed eight new RVD specificities and found that RVDs not present in natural Xanthomonas TALEs can be functional