Chitosan-DNA nanoparticles were prepared using a complex coacervation process. The important parameters for the nanoparticle synthesis were investigated, including the concentrations of DNA, chitosan and sodium sulfate, temperature of the solutions, pH of the buffer, and molecular weights of chitosan and DNA. At an amino group to phosphate group ratio (N/P ratio) between 3 and 8 and a chitosan concentration of 100 microg/ml, the size of particles was optimized to approximately 100--250 nm with a narrow distribution, with a composition of 35.6 and 64.4% by weight for DNA and chitosan, respectively. The surface charge of these particles was slightly positive with a zeta potential of +12 to +18 mV at pH lower than 6.0, and became nearly neutral at pH 7.2. The chitosan-DNA nanoparticles could partially protect the encapsulated plasmid DNA from nuclease degradation as shown by electrophoretic mobility analysis. The transfection efficiency of chitosan-DNA nanoparticles was cell-type dependent. Typically, it was three to four orders of magnitude, in relative light units, higher than background level in HEK293 cells, and two to ten times lower than that achieved by LipofectAMINE-DNA complexes. The presence of 10% fetal bovine serum did not interfere with their transfection ability. Chloroquine could be co-encapsulated in the nanoparticles at 5.2%, but with negligible enhancement effect despite the fact that chitosan only showed limited buffering capacity compared with PEI. The present study also developed three different schemes to conjugate transferrin or KNOB protein to the nanoparticle surface. The transferrin conjugation only yielded a maximum of four-fold increase in their transfection efficiency in HEK293 cells and HeLa cells, whereas KNOB conjugated nanoparticles could improve gene expression level in HeLa cells by 130-fold. Conjugation of PEG on the nanoparticles allowed lyophilization without aggregation, and without loss of bioactivity for at least 1 month in storage. The clearance of the PEGylated nanoparticles in mice following intravenous administration was slower than unmodified nanoparticles at 15 min, and with higher depositions in kidney and liver. However, no difference was observed at the 1-h time point.