Metaproteomics is an essential approach to analyse the in situ metabolic activity of microbes across various environments. In such highly diverse environmental samples, the functionality of specific microorganisms of importance often remains underexplored due to the protein inference problem arising from sequence homologies between organisms. One approach to overcome this challenge is the enrichment of non-culturable target organisms. However, this often result in samples with low protein content. In this study, we have developed a workflow that combines fluorescence in situ hybridisation (FISH) and fluorescence-activated cell sorting (FACS) with mass spectrometry-based proteomics to analyse proteins from non-culturable bacteria directly from environmental samples. We show that 1x105 cells are sufficient for reliable qualitative protein identifications, while 5x105 to 1x106 cells allow for reproducible protein quantification after FISH and FACS. Furthermore, the use of a taxon-specific database improves data analysis by significantly reducing the size of protein groups compared to metaproteomics data.