Controlling for the effect of arterial-CO2 fluctuations in resting-state fMRI: Comparing end-tidal CO2 clamping and retroactive CO2 correction

Abstract

The BOLD signal, as the basis of functional MRI, arises from both neuronal and vascular factors, with their respective contributions to resting state-fMRI still unknown. Among the factors contributing to "physiological noise", dynamic arterial CO₂ fluctuations constitutes the strongest and the most widespread modulator of the grey-matter rs-fMRI signal. Some important questions are: (1) if we were able to clamp arterial CO₂ such that fluctuations are removed, what would happen to rs-fMRI measures? (2) falling short of that, is it possible to retroactively correct for CO₂ effects with equivalent outcome? In this study 13 healthy subjects underwent two rs-fMRI acquisitions. During the "clamped" run, end-tidal CO₂ (PETCO₂) is clamped to the average PETCO₂ level of each participant, while during the "free-breathing" run, the PETCO₂ level is passively monitored but not controlled. PETCO₂ correction was applied to the free-breathing data by convolving PETCO₂ with its BOLD response function, and then regressing out the result. We computed the BOLD resting-state fluctuation amplitude (RSFA), as well as seed-independent mean functional connectivity (FC) as the weighted global brain connectivity (wGBC). Furthermore, connectivity between conditions were compared using coupled intrinsic-connectivity distribution (ICD) method. We ensured that PETCO₂ clamping did not significantly alter heart-beat and respiratory variation. We found that neither PETCO₂ clamping nor correction produced significant change in RSFA and wGBC. In terms of the ICD, PETCO₂ clamping and correction both reduced FC strength in the majority of grey matter regions, although the effect of PETCO₂ correction is considerably smaller than the effect of PETCO₂ clamping. Interestingly, we found the effect of the commonly employed white-mater/cerebrospinal-fluid regression to be similar to that of PETCO₂ clamping than global-signal regression. Nonetheless, both methods reduce functional connectivity significantly more than does PETCO₂ clamping. Furthermore, while PETCO₂ clamping reduced inter-subject variability in FC, PETCO₂ correction increased the variability. Overall PETCO₂ correction is not the equivalent of PETCO₂ clamping, although it shifts FC values towards the same direction as clamping does.