Gene correction by double strand break- or, preferably, single strand nick-induced homology-directed repair requires highly specific nucleases (e.g. engineered homing endonucleases, zinc-finger or TALE nucleases). Although engineering of such highly specific nucleases has been successful, the probability of non specific DNA cleavage (off-target cleavage) still precludes gene therapy of monogenetic diseases of humans. In an effort to develop a novel highly specific nuclease we have fused an inactive variant of the homing endonuclease I-SceI as a DNA binding module to MutH as a cleavage module. I-SceI recognizes an 18 bp sequence and is specific enough to address a unique site in the human genome. MutH is a DNA mismatch repair enzyme and is responsible for nicking the DNA at a GATC site in the vicinity to a base pair mismatch. Since MutH needs the guidance of the MutS/MutL mismatch recognition cascade to become catalytically active, the fusion enzyme MutH-ISceI should not be able to cleave random GATC sites under physiological conditions. Only if a GATC-site is addressed by an I-SceI recognition site, the fusion enzyme should be targeted to the addressed site by the binding module I-SceI, allowing MutH to catalyze a site specific nick. Our engineered MutH-ISceI constructs were tested for activity and specificity in vitro. It could be shown that the fusion construct has a clear preference for nicking the addressed target site, whereas unaddressed GATC sites remain uncleaved.