Stem cells have both self-renewing and homing/differentiative capabilities and thereby provide for life-long cell replacement in tissues and organs. They are therefore the ideal targets for most gene therapy protocols, i.e. both for long-term and transient gene expression protocols where they are either the long term carriers of the therapeutic gene (inherited/degenerative disease) or the mobilized/recruted targets of a transient regenerative process such as the formation of new blood vessels. Long-term gene therapy is thus amenable to synergistic combinations where ex vivo/in vivo genetic engineering of stem cells can be associated with in vivo transient topical expression of minigenes encoding various factors such as homing, regenerative and differentiative ones. Such a strategy is still hampered by many hurdles of which random integration of therapeutic DNA is a major safety concern. Emerging technologies are however aimed at efficient site-specific integration of therapeutic transgenes and at endogenous gene repair/modification. Promising site-specific integration vectors are at the pre-clinical stage and rely on an adeno-associated virus (AAV) rep platform or on phage phiC31 integrase. However, unlike these approaches, gene targeting is driven by homologous recombination and has thus target flexibility. It mediates DNA exchanges between chromosomal DNA and transfecting/transducing DNA, thereby providing the means to modify at will the sequence of target chromosomal DNA. Gene targeting stands thus as the ultimate process both for gene repair/alteration and targeted (i.e. site-specific) transgene integration. Such a process is however highly inefficient unless target chromosomal DNA is struck by a double-strand break (DSB). In addition, it is overwhelmed by random integration. In order to increase gene targeting frequency and eliminate random integration, we devised an approach that relies on the transfer into target cells of premade presynaptic filaments, i.e. the very stochiometric complexes between recombinase protein and single-stranded DNA (ssDNA) that mediate the key reaction from homologous recombination (homologous DNA pairing-strand exchange with double-stranded [ds] DNA). Upon publication of the enzymatic properties of recombinase RAD51, we invented chimeric presynaptic filaments with a dsDNA core, and shifted from gene conversion to 'true' gene targeting. Chimeric zinc finger nucleases are now available that create sequence-specific DSBs in target chromosomal DNA and stimulate gene targeting as expected. Such designed nucleases open exciting potentialities for standard gene targeting (non-viral) but also for emerging AAV gene targeting that has been shown to be boosted by DSB too. In these approaches, gene targeting frequency is raised to the random-integration level, i. e. ~1% of transfected/transduced cells. Our current approach is thus discussed in terms of synergistic combinations in which random-integration is blocked by the use of premade presynaptic complexes while homologous recombination is promoted both by premade presynaptic complexes and by sequence-specific DSB of target chromosomal DNA. Long-term gene therapy is thus amenable to sophisticated protocols in which stem-cell gene-targeting is combined with transient regenerative gene therapy, and might therefore apply to the treatment of the neurologic symptoms of the Lesch–Nyhan disease in which the target is our paragon model, the hypoxanthine-guanine phosphoribosyl transferase (HPRT) gene.