1 Main text

The transport, transformation, production, degradation, and mineralization of organic carbon in inland ecosystems largely determine greenhouse gas emissions and the carbon balance, with implications for climate change (Battin et al. 2009). Despite that lakes cover < 2% of the Earth's surface area, they act as a fundamental link between the carbon cycles of terrestrial and inland water ecosystems. Globally, lakes and other wetland ecosystems fix organic carbon equivalent to ~ 50% of the organic carbon uptake by the oceans (Tranvik et al. 2009). Dissolved organic matter (DOM) typically constitutes the dominant pool of organic carbon in freshwaters (Drake et al. 2023; Zhou et al. 2023). Due to the differential degradation of DOM in lakes depending on bioavailability and photodegradability, the role of lakes in offsetting human CO2 emissions is uncertain (Jia et al. 2023). DOM preserved for a long period of time is categorized as recalcitrant DOM (RDOM), considered as a carbon sink (** the distribution patterns of planktonic bacteria in lakes and species richness (Zhou et al. 2021). Higher microbial diversity may provide more metabolic pathways to degrade and produce RDOM molecules (Zhou et al. 2021). Elucidating the chemical composition and quantifying the size of the RDOM pool are highly relevant to obtain carbon neutrality. Further disentangling the coupling linkages between environmental factors, bacterioplankton, and DOM is crucial for the understanding of the stability of carbon sources and sinks in lakes (**a et al. 2022) and should be a future research priority.

There is a lack of hypotheses associated with the production and stability of RDOM in lake ecosystems. In comparison, reasons for the stability of RDOM in the oceans are currently summarized under the umbrella of three major hypotheses, the "environment hypothesis", the "intrinsic stability hypothesis", and the "molecular diversity hypothesis" (Dittmar 2015). The "environmental hypothesis" of RDOM production links DOM to the responsiveness of specific environmental conditions or to specific time periods (Dittmar 2015). These include the composition and activity of planktonic bacterial communities, redox state, mineral binding, and substrate accessibility by microorganisms and their enzymes. For example, limited availability of inorganic and organic nutrients may limit bacterial growth and DOM utilization through competition. Elucidating the key factors that influence the degradability and stability of DOM in lakes is an important area of future research. The "intrinsic stability hypothesis" links the reactivity of DOM to its molecular composition (Dittmar 2015). The molecular composition of RDOM may be synthesized by aquatic microbes, or it may be altered by abiotic modifications due to physicochemical perturbations. These abiotic modifications may result in the formation of a more stable fraction of DOM (Cory et al. 2014), thus contributing to the RDOM pool in lakes. The "molecular diversity hypothesis" suggests that bacteria in the deep ocean must maintain appropriate enzymes, transport systems, and metabolic pathways for DOM catabolism and anabolism (Dittmar 2015; Arrieta et al. 2015). However, such a strategy is difficult to have if the RDOM has a diverse chemical composition or the concentration is too low to be utilized by microbes. Due to the complexity of lake environments and the high biological activity in lakes, the composition of DOM and the stability of the molecular structure vary tremendously among different aquatic organisms in lakes. The "environmental hypothesis" mechanism may play an important role in the stability of DOM in lake ecosystems. Therefore, identifying the interactions between RDOM and planktonic bacteria across different environmental gradients (e.g., trophic status, catchment land use, sorption to metals and clay minerals) is key to understanding the dynamics of RDOM in lakes.

Fortunately, advances are made in metagenomic analysis and DOM identification, involving growing utilization of high- and ultrahigh-resolution mass spectrometry (e.g., FT-ICR MS and Orbitrap MS) (Hu et al. 2022). This provides unique opportunities to explore the link between microbial carbon pumps and the mechanisms of RDOM formation and to deepen our understanding of the interactions between the microbial community and the chemical composition of DOM as well as the outgassing of greenhouse gases from lakes (Li et al. 2022; McDonough et al. 2022). Big data and systems science approaches, including machine learning and numerical simulation, can further help to elucidate the role of RDOM in the lake carbon cycling and the role of lakes relative to the future climate change.