Dbx1 pre-Bötzinger complex interneurons comprise the core inspiratory oscillator for breathing in adult mice

The brainstem pre-Bötzinger complex (preBötC) generates inspiratory breathing rhythms, but which neurons comprise its rhythmogenic core? Dbx1-derived neurons may play the preeminent role in rhythm generation, an idea well founded at perinatal stages of development but not in adulthood. We expressed archaerhodopsin or channelrhodopsin in Dbx1 preBötC neurons in intact adult mice to interrogate their function. Prolonged photoinhibition slowed down or stopped breathing, whereas prolonged photostimulation sped up breathing. Brief inspiratory-phase photoinhibition evoked the next breath earlier than expected, whereas brief expiratory-phase photoinhibition delayed the subsequent breath. Conversely, brief inspiratory-phase photostimulation increased inspiratory duration and delayed the subsequent breath, whereas brief expiratory-phase photostimulation evoked the next breath earlier than expected. Because they govern the frequency and precise timing of breaths in awake adult mice with sensorimotor feedback intact, Dbx1 preBötC neurons constitute an essential core component of the inspiratory oscillator, knowledge directly relevant to human health and physiology.

Here we reevaluate the inspiratory rhythmogenic role of Dbx1 preBötC neurons in adult mice 71 with intact sensorimotor feedback. Using optogenetic technologies to photoinhibit or 72 photostimulate Dbx1 neurons, we show that their perturbation affects breathing frequency and 73 the precise timing of individual breaths within the breathing cycle, which are key properties of a 74 core oscillator microcircuit. Other respiratory and non-respiratory roles notwithstanding, these 75 data indicate that Dbx1 preBötC neurons constitute an essential core oscillator for inspiration. 76 subsequent inspiration occurred earlier than expected (Φ Shift = -147 ± 23º, p = 1e-6, n = 4) while 136 shortening inspiratory time (T i ) by almost half (∆T i = 45 ± 5%, p = 1e-6, n = 4) ( Figure 3A 1,2 and 137 A 3 top trace). Brief photoinhibition also evoked significant phase advances and reduced T i 138 during the rest of inspiration (Φ Stim of 30-120º), but the magnitude of those changes 139 monotonically decreased as Φ Stim approached the inspiratory-expiratory transition. 140 Brief photoinhibition did not perturb the system during the inspiratory-expiratory transition (Φ Stim 141 of 120-180º). During early expiration (Φ Stim of 180-210º), which is often referred to as post-142 inspiration (Anderson et al., 2016;Dutschmann et al., 2014) we observed the first significant 143 phase delay such that the subsequent inspiration occurred later than expected in response to 144 brief photoinhibition (Φ Shift = 32 ± 7º, p = 0.006, n = 4, Figure 3A 1 and A 3 bottom trace). Phase 145 delays were consistently evoked during expiration (Φ Stim of 210-360º) with a maximum phase 146 delay during late expiration (Φ Stim of 300-330º) (Φ Shift = 78 ± 10º, p = 1e-6, n = 4). Brief 147 photoinhibition during expiration did not affect T i , which is a straightforward result because the 148 inspiratory period had ended ( Figure 3A 2 ). Note, that ∆T i was statistically significant at Φ Stim of 149 210-240º) but that change is not physiologically meaningful because the magnitude of the 150 change is small and not part of a consistent trend in the phase-response curve. 151 The relationship between Φ Stim and the phase of the subsequent breath (Φ N+1 , Figure 3  In contrast to its effects on breathing phase (Φ Shift and Φ N+1 ), brief photoinhibition had little effect 155 on V T throughout most of the respiratory cycle with changes of less than 10% across the entire 156 respiratory cycle, except during early inspiration (Φ Stim of 0-30º, in which V T decreased by 23 ± 157 8%, p = 0.02, n = 4) and early expiration (Φ Stim of 150-180º, in which V T increased by 16 ± 11%, 158 p = 0.01, n = 4) ( Photoinhibition during early inspiration (Φ Stim of 0-30º) caused a phase advance (Φ Shift = -86 ± 166 16º, p = 1e-5, n = 4). The first significant phase delay in the awake animal occurred when brief 167 photoinhibition was applied during peak expiration (Φ Stim of 210-240º, Φ Shift = 68 ± 15º, p = 1e-6, 168 n = 4). Φ Shift tended to increase as brief photoinhibition was applied at later points during the 169 expiratory phase. The maximum phase delay occurred during late expiration (Φ Stim of 330-360º, 170 Φ Shift = 118 ± 25º, p = 4e-5, n = 4) ( Figure 3B 1 and B 3 ). Brief photoinhibition decreased T i by 171 nearly one-third (∆T i = 28 ± 9%, p = 1e-5, n = 4) during early inspiration (Φ Stim of 0-30º) but had 172 no significant effect at any other time during the cycle. 173

Photostimulation of Dbx1 preBötC neurons enhances breathing and modifies the timing 174 and magnitude of breaths 175
We illuminated the preBötC unilaterally in sedated adult Dbx1;CatCh mice (the intersection of a 176 Dbx1 CreERT2 driver mouse and a reporter featuring Cre-and Flp-dependent calcium translocating 177 channelrhodopsin [CatCh] expression) following viral transduction in the preBötC with a 178 synapsin-driven Flp recombinase. Using this double-stop intersectional approach, CatCh-EYFP 179 expression was limited to the preBötC ( Figure 1D). In control conditions ƒ was typically ~3 Hz, 180 V T was ~0.1 ml, and MV was ~50 ml/min. Bouts of blue light (473 nm) at three intensities 181 significantly increased ƒ by 0.8, 1.1, and 1.3 Hz, respectively (t-test, p = 0.03, 0.005, and 0.03, n 182 = 4). There were no significant effects on V T or MV at any light intensity ( Figure 4A and B). 183 We repeated these unilateral photostimulation experiments in Dbx1;CatCh mice while awake 184 and unrestrained. Frequency increased by 1.6 Hz in response to light at the highest intensity 185 ( Figure 4C and D). There were no other notable changes in ƒ, V T , or MV at any light intensity. 186 In wild type littermates, we observed no effects on breathing in either sedated or awake mice in 187 response to light at any intensity (Figure 4 -figure supplement 1). 188 Therefore, these data collectively show that CatCh-mediated photostimulation of Dbx1 preBötC 189 neurons selectively enhances breathing frequency in sedated and awake intact mice. 190 Next we applied brief (100 ms) light pulses at different time points during the breathing cycle. 191 Unilateral illumination of the preBötC during inspiration caused a phase delay and increased T i . 192 The maximum phase delay occurred during peak inspiration (Φ Stim of 60-90º, Φ Shift = 125 ± 18º, p 193   period. Brief photostimulation during inspiration prolonged it (i.e., increased Ti) and delayed the 274 next cycle (i.e., a phase delay). The straightforward interpretation is that CatCh-mediated inward 275 current augments recurrent excitation thus prolonging inspiratory burst duration. Overexcited 276 rhythmogenic neurons require more time to recover, which lengthens cycle time and delays the 277 subsequent inspiration. 278 We observed that photostimulation at any other point in the cycle evoked inspiration earlier than 279 expected, a phase advance, but did not otherwise modify inspiration. In contrast to a prior 280 report, brief photostimulation did not evoke phase advances during early expiration (Alsahafi et  Brief photoinhibition of Dbx1 preBötC neurons during inspiration shortened it (i.e., decreased T i ) 302 and initiated the next cycle earlier than expected, a phase advance. We infer that 303 hyperpolarizing rhythmogenic neurons checks the recurrent excitation process, which impedes 304 but does not prevent inspiration. Nevertheless, the evoked breath is shorter in duration. preBötC 305 neurons do not overexcite or become refractory, which facilitates the onset of the next cycle, 306 hence the phase advance. That mechanism, here evoked by ArchT, mirrors the role of 307 endogenous phasic synaptic inhibition, which curbs recurrent excitation to limiting inspiratory 308 activity and facilitate inspiratory-expiratory phase transition (Baertsch et al., 2018). We found 309 that photoinhibition during expiration consistently caused a phase delay, which indicates 310 hyperpolarization of Dbx1 preBötC neurons resets recurrent excitation and thus prolongs the 311 interval until the next inspiration. We implanted fiber optic appliances bilaterally in Dbx1;ArchT mice and unilaterally in 484 Dbx1;CatCh mice at a depth of 5.5 to 5.9 mm from bregma, which were secured with a 485 cyanoacrylate adhesive (Loctite 3092, Henkel Corp., Rocky Hill, CT, USA). Dbx1;ArchT animals 486 recovered for a minimum of 10 days before any further experimentation. Dbx1;CatCh mice 487 recovered for a minimum of 21 days before further experimentation. 488

Breathing measurements 489
After anesthetizing mice using 2% isoflurane we connected the ferrules of Dbx1;ArchT mice to a 490 589-nm laser (Dragon Lasers, Changchun, China). The ferrule of Dbx1;CatCh mice was 491 connected to a 473-nm laser (Dragon Lasers). Mice recovered from isofluorane anesthesia for 492 ~1 hr, and then we measured breathing behavior using a whole body plethysmograph (Emka 493 Technologies, Falls Church, VA, USA) that allowed for fiberoptic illumination in a sealed 494 chamber. 495 In a separate session, these same mice were lightly sedated via intraperitoneal ketamine 496 injections (15 mg/kg minimum dose), which we titrated as needed to reduce limb movements 497 but retain toe-pinch and blink reflexes. The maximum aggregate dose was limited to 50 mg/kg. 498 Mice were fitted with a modified anesthesia mask (Kent Scientific, Torrington, CT, USA) to 499 measure breathing. 500 We applied a circuit of positive pressure, with balanced vacuum, to continuously flush the 501 plethysmograph with breathing air. The plethysmograph and the mask were connected to a 1-502 liter respiratory flow head and differential pressure transducer that measured airflow; positive 503 airflow reflects inspiration in all cases. Analog breathing signals were digitized at 1 kHz 504 (PowerLab). 505

Optogenetic protocols 506
We applied 5 s bouts of light (either 473 or 589 nm) to Dbx1;ArchT and Dbx1;CatCh mice at 507 graded intensities of 6.8, 8.6, and 10.2 mW. All ferrules were tested with a power meter prior to 508 implantation to verify that illumination intensity did not vary more than 0.1 mW from the specified 509 values. Bouts of light application were separated by a minimum interval of 30 s. We also applied 510 100 ms light pulses at a fixed intensity of 10.2 mW. We exposed each mouse to 85-200 pulses 511 spaced at random intervals of between 1 and 5 s. 512 We applied 2 s bouts of 589-nm light (at the same intensities listed above) to rhythmically active 513 slices. The fiberoptics were targeted to selectively illuminate the preBötC bilaterally but not the 514 adjacent reticular formation. 515

Data analyses 516
The airflow signal was band-pass filtered (0.1-20 Hz) and analyzed using LabChart 8 software 517 (AD Instruments), which computes airflow (units of ml/s), respiratory rate (i.e., frequency, ƒ, 518 units of Hz), tidal volume (VT, units of ml), inspiratory time (T i ), and minute ventilation (MV, units 519 of ml/min). We computed statistics using GraphPad Prism 6 (La Jolla, CA, USA) and R: The 520 Project for Statistical Computing (R, The R Foundation, Vienna, Austria) and prepared figures 521 using Adobe Illustrator (Adobe Systems Inc., San Jose, CA, USA), GraphPad Prism 6, and 522 IGOR Pro 6 (Wavemetrics, Lake Oswego, OR, USA). We analyzed the experiments in which 5 s 523 light pulses were applied to the preBötC using paired t-tests, specifically comparing mean ƒ, V T , 524 and MV for control and illumination conditions at three different light intensity levels (i.e., at each 525 laser strength tested, the pre-illumination ventilation serves as its own control). 526 We analyzed phase-response relationships of the breathing cycles perturbed by 100 ms-527 duration light pulses (see Figure 3C inset). The expected cycle period was measured from the 528 unperturbed cycle immediately before the light pulse, which was defined as spanning 0-360º 529 (Φ Expected ). Cycle times were measured from the start of inspiration in one breath to the start of 530 inspiration of the subsequent breath. For perturbed cycles, 100-ms light pulses were applied at 531 random time points spanning the inspiration and expiration to test for phase shifts. Φ Stim marks 532 the phase at which the light pulse occurred. The induced cycle period (Φ Induced ) was measured 533 from the perturbed cycle. The perturbation of breathing phase, Φ Shift , was defined as the 534 difference between Φ Induced and Φ Expected . We calculated change in V T and T i in the perturbed 535 breath compared to the expected breath normalized to the expected breath (refered to as, ∆V T 536 and ∆T i , respectively). Further, we calculated the phase shift of the breath following the 537 perturbed breath (i.e., the cycle after Φ Induced ) also with respect to Φ Expected ; we refer to the phase 538 of the subsequent breath Φ N+1 . Measurements of Φ Shift , ∆V T , ∆T i , and Φ N+1 are all linked to a 539 particular Φ Stim within the interval 0-360º. To analyze group data we sorted Φ Stim into 12 equally 540 sized 30º bins. We computed the mean and standard deviation (SD) for Φ Shift , ∆V T , ∆T i , and Φ N+1 541 within each bin, which we then plotted in phase-response curves along with values calculated 542 from wild type littermates. A Tukey's HSD to test was used to evaluate how unlikely it would 543 have been to obtain mean Φ Shift , ∆V T , ∆T i , and Φ N+1 for each bin if the optogenetic perturbations 544 had commensurate effects on Dbx1;ArchT (or Dbx1;CatCh) mice and wild type littermates.   Dbx1;CatCh mice (n = 4, cyan) and wild type littermates (n = 4, magenta). A 1 , Phase-response 845 curve plotting Φ Shift following 100-ms photostimulation at Φ Stim throughout the breathing cycle in 846 sedated mice. Φ Stim was partitioned into 12 equally sized bins (30º) in A and B. A 2 , Phase-847 response curve for changes in T i following photostimulation (i.e., the perturbed breath) in the 848 same cohort of sedated mice. The abscissa marks the inspiratory (I, 0-150º) and expiratory (E, 849 150-360º) phases of the breathing cycle (0-360º), which applies to A 1 and A 2 . A 3 , Sample airflow 850 traces from a representative sedated mouse (Φ Stim is indicated by an orange bar and numeral 851 value). Time calibration as shown. B 1 , Phase-response curve plotting Φ Shift following brief 852 photostimulation at Φ Stim throughout the breathing cycle in awake unrestrained mice. B 2 , Phase-853 response curve for changes in T i following brief photostimulation (i.e., the perturbed breath) in 854 the same cohort of awake unrestrained mice. The abscissa marks the inspiratory (I, 0-110º) and 855 expiratory (E, 110-360º) phases of the complete breathing cycle (0-360º), which applies to B 1 856 and B 2 . B 3 , Sample airflow traces from a representative awake unrestrained mouse (Φ Stim is 857 indicated by an orange bar and numeral value). Time calibration is shown. 858 Dbx1;CatCh mice (n = 4, cyan) or wild type littermates (n = 4, magenta). A 1 , Phase-response 861 curve plotting Φ N+1 vs. Φ Stim throughout the breathing cycle in sedated mice. A 2 , Phase-862 response curve for changes in V T following photostimulation (i.e., the perturbed breath) in the 863 same cohort of sedated mice (n = 4). The abscissa marks the inspiratory (I, 0-150º) and 864 expiratory (E, 150-360º) phases of the breathing cycle (0-360º), which applies to A 1 and A 2 . B 1 , 865 Phase-response curve plotting Φ N+1 vs. Φ Stim in awake unrestrained mice. B 2 , Phase-response 866 curve for ∆V T vs. Φ Stim in the same cohort of awake unrestrained mice. The abscissa marks the 867 inspiratory (I, 0-110º) and expiratory (E, 110-360º) phases of the complete breathing cycle (0-868 360º), which applies to B 1 and B 2 . 869