Sara van den Berg

157 6 T-cell dynamics in CMV selected CMV+ individuals with high CD8 + CMV-specific T-cell numbers, we did not find any kinetic differences between CMV-specific and bulk memory T-cells. We do not know whether any CMV reactivations occurred during follow-up in our participants, and cannot exclude the possibility that in these extremely healthy individuals CMV reactivation may occur less frequently than in individuals with an impaired health, which may in turn influence the dynamics of the CD8 + T-cell pool. Thanks to the rather strict selection criteria, we were able to study the influence of CMV infection on the T-cell pool in the absence of other underlying (chronic) conditions. To understand how T-cell memory is maintained, we investigated the production and loss rates of cell populations in vivo in humans and put this in context of the expression of proliferation, apoptosis, and senescence markers. Generally speaking, expression of senescence markers as well as Ki-67 correlated with T-cell production rates in vivo . CMV-specific CD8 + T-cells had on average a higher expression of senescence markers than their bulk memory counterparts, though this was not reflected in differences in in vivo T-cell production rates. This might be partly explained by the fact that we compared the in vivo production rates of total CD8 + CMV- specific T-cells to that of bulk T EM/EMRA CD8 + T-cells, a comparison for which the expression of Ki-67 also did not differ significantly. Deuterium labeling studies tend to be limited to relatively small numbers of participants, which makes it more difficult to draw firm conclusions on kinetic comparisons between groups with a lot of interindividual variation. Nevertheless, deuterium labelling provides a powerful tool to quantify cellular dynamics, not in the least because kinetic estimates are based on multiple data points per individual, making the estimates per individual less sensitive to fluctuations over time. As a result, deuterium labelling studies, even when based on relatively small numbers of participants, can yield very reliable and reproducible results. We and others have for example reproducibly shown that CD4 + T-cells have higher production rates than CD8 + T-cells [43]. Furthermore, we used the kinetic heterogeneity model as presented here [28] and the multi-compartment model (data not shown) [29]. When using the latter, we also did not find differences in production rates between CD8 + CMV-specific T-cells and bulk CD8 + T EM/EMRA cells. Even when estimating production rates by simple linear regression on the data during the uplabeling period, the conclusion on the effect of CMV did not change ( Supplementary Figure 9A,B ). However, our data might suggest a need for new mathematical models that can account for smaller differences in label uptake without overfitting the data. Since the estimations of cellular loss rates relied on a combination of deuterium labelling data and measurements of cell frequencies or numbers, which are notoriously fluctuating, these estimates are probably less precise than those of the production rates. Differences therein could in theory still bridge the discrepancy between an increased late-stage differentiation state and a similar production rate. We thus cannot completely rule out that CMV-specific T-cells might be slightly longer lived. A downside of deuterium labelling is that it cannot be used to study single-cell kinetics and needs to be performed on sorted populations based on predefined cell surface markers. It could be that kinetically different subpopulations cannot be subdivided along the same lines as the well-established memory T-cell populations (e.g. T CM , T EM , T EMRA ), and that other