The role of the sputum microbiome in COPD progression remains elusive. As the advent of the new culture-independent microbial sequencing technique makes it possible to disclose the complex microbiome community of the respiratory tract. So we used Metagenomic Next-Generation Sequencing(mNGS) to confirm whether there are differences in sputum microbiome of COPD between different exacerbation frequencies and lung function. 39 COPD patients were divided into a frequent exacerbators (FE) group(n=20) and a non-frequent exacerbators (NFE)(n=19) group according to their exacerbation history, or a mild group (FEV1/pre?50%, n=20)and a severe group(FEV1/pre<50%, n=19) . Sputum was collected during their stable phase, followed by DNA extraction, untargeted mNGS and bioinformatic analysis. mNGS identified 3355 bacteria, 71 viruses and 22 fungi at the specie level. It was found that Shannon index and Simpson index in FE group was lower than that in NFE group(p=0.005, 0.008,respectively) but similar between mild and severe groups. Out of top 10 bacteria taxa, Veillonella, Fusobacterium and Prevotella jejuni had a higher abundance in NFE group, Rothia had a higher abundance in mild group. Linear discriminant analysis revealed that many bacterial taxa were more abundant in NFE group, and they mostly belonged to Actinobacteria, Bacteroidetes and Fusobacteria phyla. Frequency of exacerbations was also found to be negatively correlated with alpha diversity (with Shannon index, r=-0.423; with Simpson index, r=-0.482). These findings demonstrated that sputum microbiome community dysbiosis was associated with different exacerbation frequencies and lung function in stable COPD.