Fig. 7. Gasps collapse latent orbits into ballistic, all-or-none efforts.
a, Integrated diaphragm activity illustrating the transition to gasping during hypoxia presentation. b, Example spike raster and diaphragm activity during a period of gasping. Approximate rostral–caudal boundaries of some landmark medullary regions are shown. c, Breath onset-aligned average firing rates of two example units. Maximum firing rate (FR) of top unit increases and bottom unit decreases during gasping. Shaded region is mean FR ± s.e.m. sp/s, spikes/second. d, Firing rates of all units recorded in b for eupnea (x axis) and gasps (y axis). Each dot is a recorded unit. Color indicates preferred phase of firing during eupnea: inspiratory (red), expiratory (blue) or tonic (gray). e, Breath onset-aligned average of the leading three PCs in eupnea (gray) and gasps (brown) for the recording in b. f, Population trajectories through PC space for the recording in b for eupnea (gray) and gasps (brown). g, Trajectory speed (radial axis, a.u.) as a function of breathing phase (angular axis) for eupnea (gray) and gasps (brown). Top of circle is inspiratory effort; bottom of circle is IBI. Shaded region is mean ± s.e.m. across all recordings. h, Real and imaginary (Imag.) components of the eigenvalues of the dynamics matrices that govern the inspiratory and expiratory states in the rSLDS models fit during gasping. Only recordings in which the dynamics were regenerative during control eupnea (n = 11) are shown. Dashed line is the unit circle. i, Correlation (Corr.) values for all pairs of units during eupnea in O2, room air, and during gasps. Red is negative correlations; black is positive correlations. j, Correlation value of each pair of units during O2 presentation (x axis) against those during room air presentation (y axis, left) and during gasps (y axis, right). k, Linear fit (R value) of the pairwise correlation values in O2 compared to room air and gasps. Each dot is a recording; error bars are 95% confidence interval. We consider room air as a control change in correlation structure as compared to O2. Changes in population correlation structure during sigh are different than those observed in room air (two-sided Mann–Whitney U test ***P = 7.3 × 10−14, nrecs_roomair = 116, nrecs_gasp = 38).