ABSTRACT: BACKGROUND: Chromatin immunoprecipitation coupled with high-throughput DNA sequencing (ChIP-seq) offers high resolution, genome-wide analysis of DNA-protein interactions. However, current standard methods require abundant starting material in the range of 1--20 million cells per immunoprecipitation, and remain a bottleneck to the acquisition of biologically relevant epigenetic data. Using a ChIP-seq protocol optimised for low cell numbers (down to 100,000 cells / IP), we examined the performance of the ChIP-seq technique on a series of decreasing cell numbers. RESULTS: We present an enhanced native ChIP-seq method tailored to low cell numbers that represents a 200-fold reduction in input requirements over existing protocols. The protocol was tested over a range of starting cell numbers covering three orders of magnitude, enabling determination of the lower limit of the technique. At low input cell numbers, increased levels of unmapped and duplicate reads reduce the number of unique reads generated, and can drive up sequencing costs and affect sensitivity if ChIP is attempted from too few cells. CONCLUSIONS: The optimised method presented here considerably reduces the input requirements for performing native ChIP-seq. It extends the applicability of the technique to isolated primary cells and rare cell populations (e.g. biobank samples, stem cells), and in many cases will alleviate the need for cell culture and any associated alteration of epigenetic marks. However, this study highlights a challenge inherent to ChIP-seq from low cell numbers: as cell input numbers fall, levels of unmapped sequence reads and PCR-generated duplicate reads rise. We discuss a number of solutions to overcome the effects of reducing cell number that may aid further improvements to ChIP performance.