The aquatic environments in the two Antarctic Long Term Ecological Research sites, Palmer (PAL) and the McMurdo Dry Valleys (MCM), represent starkly contrasting polar ecosystems. PAL, a seasonally ice free marine environment, spans from highly productive nearshore waters harboring a rich guild of large predators to less productive waters over the continental slope. MCM covers a range of aquatic and terrestrial environments including several perennially ice covered lakes of marine origin. These lakes are characterized by their isolation, permanent ice-covers, chemical stratification, and few metazoan predators. Despite these differences the trophic structure of PAL and MCM are influenced by many of the same environmental factors, including low temperature and changing climate. We leveraged the ecological differences between these two sites as a natural experiment, comparing the long term records of bacterial production (BP), primary production (PP), and dissolved organic carbon (DOC) concentration to identify ecosystem processes that may represent fundamental differences between sites. Using generalized additive and generalized least-squares models we identified significant temporal trends in the ratio of BP:PP and the concentration of DOC at both PAL and MCM. Surprisingly, the direction of the trend was different between the two sites, with DOC concentration increasing at PAL and decreasing within the photic zone of Lake Fryxell and the west lobe of Lake Bonney. The dynamics of the relationship between BP and PP are variable between sites across the same time period, suggesting that bacterial heterotrophy and the DOC pool are coupled in fundamentally different ways. To develop further hypothesis regarding microbial community function we used data from the MIRADA-LTERS project to infer the likely metabolisms of the bacterial community at PAL and in Lakes Fryxell and Bonney. We found that low abundance metabolic pathways, many of which function in energy and carbon acquisition, accounted for much of the variance between samples from PAL and MCM. Taxonomic differences, by contrast, were primarily due to the distribution of relatively abundant taxa. We hypothesize that functional differences between the MCM lakes and PAL marine communities are responsible for the variable relationship between BP, PP, and DOC across sites. Time series features observed at all sites, such as a low BP:PP anomaly in 2010, may be driven by continent-wide climate events such as the strong amplitude of the Southern Annual Mode (SAM) in 2009-2010.