Recent Arctic studies suggest that sea-ice decline and permafrost thawing will affect phytoplankton dynamics and stimulate marine heterotrophic communities. However, in what way the plankton composition will change as the warming proceeds remains elusive. Here we investigate the chemical signature and plankton speciation of the supra-micron (> 10 μm) particulate organic matter (supra-POM) fraction collected along the Siberian shelf. Supra-POM samples were analysed at bulk (<i>δ</i><sup>13</sup>C and Δ<sup>14</sup>C) and molecular level (CuO oxidation and IP<sub>25</sub>) while plankton identification established the dominant taxa. In addition, surface water chemical properties were integrated with the plankton dataset to understand the link between plankton composition and environmental conditions. <br><br> The dual-carbon isotope fingerprint indicates a large variability in the supra-POM distribution while terrestrial biomarkers suggest negligible land-derived input. In the open-waters of the outer Laptev Sea (LS), heterotrophic plankton dominated the assemblages. <i>δ</i><sup>13</sup>C and Δ<sup>14</sup>C suggest that modern terrestrial dissolved organic carbon (DOC) from the Lena river is the primary source of metabolizable carbon which is transferred to the heterotrophic communities via microbial loops. Moving eastwards toward the sea-ice dominated East Siberian Sea (ESS), the system became progressively more autotrophic and dominated by sea-ice and pelagic diatoms which is confirmed. Comparison between <i>δ</i><sup>13</sup>C of supra-POM samples and CO<sub>2</sub>aq concentrations suggests that the carbon isotope fractionation follows the general growth vs CO<sub>2</sub>aq supply model with the highest <i>δ</i><sup>13</sup>C values found in the easternmost, most productive stations. <br><br> In a warming scenario characterized by enhanced terrestrial release and further sea-ice decline, heterotrophic conditions fuelled by terrestrial DOC will likely persist in the LS while ESS might experience enhanced primary productivity. This will result in a sharp compositional gradient similar to what documented in our semi-synoptic study.