In this paper, we present a series of heteronuclear NMR experiments for the direct observation and characterization of lysine NH3 groups in proteins. In the context of the HoxD9 homeodomain bound specifically to DNA we were able to directly observe three cross-peaks, arising from lysine NH3 groups, with 15N chemical shifts around approximately 33 ppm at pH 5.8 and 35 degrees C. Measurement of water-exchange rates and various types of 15N transverse relaxation rates for these NH3 groups, reveals that rapid water exchange dominates the 15N relaxation for antiphase coherence with respect to 1H through scalar relaxation of the second kind. As a consequence of this phenomenon, 15N line shapes of NH3 signals in a conventional 1H-15N heteronuclear single quantum coherence (HSQC) correlation experiment are much broader than those of backbone amide groups. A 2D 1H-15N correlation experiment that exclusively observes in-phase 15N transverse coherence (termed HISQC for heteronuclear in-phase single quantum coherence spectroscopy) is independent of scalar relaxation in the t(1) (15N) time domain and as a result exhibits strikingly sharper 15N line shapes and higher intensities for NH3 cross-peaks than either HSQC or heteronuclear multiple quantum coherence (HMQC) correlation experiments. Coherence transfer through the relatively small J-coupling between 15Nzeta and 13Cepsilon (4.7-5.0 Hz) can be achieved with high efficiency by maintaining in-phase 15N coherence owing to its slow relaxation. With the use of a suite of triple resonance experiments based on the same design principles as the HISQC, all the NH3 cross-peaks observed in the HISQC spectrum could be assigned to lysines that directly interact with DNA phosphate groups. Selective observation of functional NH3 groups is feasible because of hydrogen bonding or salt bridges that protect them from rapid water exchange. Finally, we consider the potential use of lysine NH3 groups as an alternative probe for larger systems as illustrated by data obtained on the 128-kDa enzyme I dimer.