Previous studies suggest that, during the late Pleistocene ice ages, surface-deep exchange was somehow weakened in the Southern Ocean's Antarctic Zone (AZ), reducing the leakage of deeply sequestered CO2 and thus contributing to the lower atmospheric CO2 levels of the ice ages. To better understand the surface nutrient consumption in the Antarctic ocean during glacial intervals and its implication on atmospheric CO2 changes, we measured diatom-bound nitrogen isotopes (d15N_db) extending back to 150 thousand years ago (150ka) in two sediment cores in the Indian sector of the AZ. The data series include d15N_db, TEX86L-based SST, d18O of planktonic foraminifera Neogloboquadrina pachyderma (sin.), sediment Ba/Fe and the age model of MD12-3394 (48°23' S, 64°35' E, 2320m water depth), d15N_db and updated age model of MD11-3353 (50°34' S, 68°23' E, 1568m water depth), as well as the mean d15N_db of the two sediment cores, and the d15N_db offset calculated using the mean d15N_db and MD12-3394 d15N_db alone. The age model for MD12-3394 was based on 7 Holocene radiocarbon dates measured with planktonic foraminifera Neogloboquadrina pachyderma (sin.) and Globigerina bulloides hand-picked from the > 63µm fraction, and correlation of reconstructed SST by TEX86L to the Antarctic temperature stack (ATS) (adjusted to AICC2012 age scale) compiled from ice core data. Ages were linearly interpolated between the stratigraphic tie points. The age model of MD11-3353 is based on the age model published in Thöle et al. (2019), with modifications of the youngest age tie point and the tie point at Marine Isotope Stage (MIS) 5-6 transition, and two additional tie points during MIS 3 based on correlation of d15N_db to MD12-3394 d15N_db. Monte Carlo simulation and Kalman filter were combined to generate a mean d15N_db time series for the two sediment cores on a 500-year time grid, following the method in Wang et al. (2017). For the d15N_db offset, a linear fit was first applied to glacial (20-27 ka,140-155 ka) and interglacial (0-10 ka, 115-126 ka) extreme values of d15N_db (or calculated mean d15N_db) and ATS. Using this linear regression, we calculated the ATS-predicted d15N_db. The d15N_db offset was obtained by subtracting the ATS-predicted d15N_db from measured d15N_db (or calculated mean d15N_db).