Control of Breathing, Altitude and Diving

Hold your breath for as long as you can and the urge that eventually defeats you comes mostly from carbon dioxide piling up rather than from oxygen running out, though falling oxygen and other signals contribute. That is an odd way to build a controller for an organ that delivers oxygen. This lesson takes it apart: where the rhythm comes from, what chemistry sets its size, and what happens to the loop when you carry it up a mountain or put your face in cold water. Gas exchange and breathing already showed that carbon dioxide is the resting stimulus; here you find where it is sensed and what it is worth in millimetres of mercury.

The rhythm is made in the medulla, its size is set by chemistry

Two questions sit inside the phrase “control of breathing”, and keeping them apart saves much confusion. One is where the alternation of inspiration and expiration comes from. The other is what sets how hard it drives the muscles of lung mechanics, the diaphragm and intercostals.

The rhythm comes from the ventral respiratory column, a longitudinal strip of the medulla. Its rostral and caudal ventral respiratory groups contain inspiratory and expiratory premotor neurons respectively, which relay drive onward to the motor neurons. Rhythm generation itself is far more local, and it is carried by coupled oscillators rather than one site: the pre-Botzinger complex generates inspiration, while a parafacial oscillator near the retrotrapezoid nucleus is implicated in active expiration. Smith and colleagues’ 1991 experiment, in a neonatal rat brainstem and spinal cord kept working in vitro, removed tissue in serial transverse microsections while recording respiratory motor output. The output survived the loss of tissue rostral and caudal to one small ventrolateral region, stopped when that region came out, and continued in thin slices containing it. They named it the pre-Botzinger complex.

Keep the scope of that result. It shows local necessity in a reduced neonatal rodent preparation, and that a slice containing the region suffices. It does not establish the adult human circuit, nor that one intrinsic pacemaker mechanism accounts for every breath. How any of these neurons fires is the action potential’s business.

Drive is a separate matter, and at rest it is largely chemical. At rest, central and peripheral respiratory chemoreflexes act together to stabilize arterial PCO2 and pH. The rest of the lesson is those two arms.

Carbon dioxide is sensed as hydrogen ions, inside the brain

Carbon dioxide is lipid-soluble and crosses the blood-brain barrier readily. Once in brain extracellular fluid and cerebrospinal fluid it is hydrated and releases hydrogen ions, and brain carbon dioxide is sensed primarily through those hydrogen ion changes rather than as the molecule itself. That fact explains why a controller tuned to carbon dioxide is also a controller of acid-base status: ventilation sets arterial PCO2, and arterial PCO2 is one of the two terms that set pH, as acid-base balance works out in full.

Which cells do the sensing is a live question. In mice, neurons of the retrotrapezoid nucleus, a small population on the ventral medullary surface, are proton-sensitive through the potassium channel TASK-2 and the proton-activated receptor GPR4, which is a G protein-coupled receptor of the kind GPCRs and G proteins covers. In rat and mouse networks, retrotrapezoid neurons project to the pre-Botzinger complex, the nucleus tractus solitarius and pontine nuclei: the route from a pH signal to a bigger breath.

As of a 2023 review by Gonye and Bayliss, several proposed central chemoreceptor cell types, astrocytes among them, had not been shown to satisfy every criterion that review set out for calling a cell a respiratory chemoreceptor. Showing that a cell is acid-sensitive and that manipulating it changes breathing is weaker than showing it is the sensor.

The carotid body is the oxygen sensor, and it is fast

The peripheral arm is a set of small, heavily perfused organs beside the great arteries, which is why it can respond within seconds. The carotid body is a small mass of tissue, on the order of ten milligrams in an adult human, at the fork of each carotid artery. It is the main peripheral chemoreceptor for arterial oxygen, carbon dioxide and pH, and its chemoreceptive glomus cells transmit hypoxic signals through the carotid sinus branch of the glossopharyngeal nerve. Aortic bodies contribute too, with oxygen-sensing afferents in the vagus; in anesthetized, paralyzed, artificially ventilated cats, aortic chemoreceptor responses to carbon dioxide were smaller than carotid responses.

Two structures at the same fork cause more exam errors than anything else here. The carotid body is a chemoreceptor and responds to arterial chemistry. The carotid sinus, a stretch-sensitive widening of the arterial wall, is a baroreceptor and responds to pressure; it belongs to regulating arterial pressure. Same bifurcation, same nerve, different organ and stimulus.

How much of the human hypoxic response does this organ carry? Timmers and colleagues’ 2003 review reported that bilateral human carotid chemoreceptor denervation permanently abolishes the normocapnic ventilatory response to hypoxia and reduces the response to hypercapnia. For mechanism underneath that, Macias and colleagues deleted Epas1, which encodes the oxygen-sensitive transcription factor HIF-2alpha, embryonically in mouse catecholaminergic tissue using tyrosine hydroxylase-driven Cre. HIF and its oxygen-dependent degradation, with the work of Semenza, Kaelin and Ratcliffe, belong to ubiquitin and proteolysis. Adult knockouts had virtually no tyrosine hydroxylase-positive glomus cells, and developing knockout carotid bodies showed a progressive rise in TUNEL-positive cells at embryonic day 18.5 and at birth. The response over the first five minutes of 10 percent inspired oxygen was substantially reduced, and in a separate measurement in isoflurane-anesthetized mice, saturation after five minutes of 10 percent oxygen was 50 to 55 percent against 73 percent in littermate controls. The response to 5 percent inspired carbon dioxide was retained, and the matching HIF-1alpha deletion kept its hypoxic response. The design problem below asks what that cannot settle.

Read the response curves, and say what was clamped

A chemoreflex is measured as a curve, and a curve without its conditions is not a measurement. Put ventilation on the vertical axis and the gas tension you are changing on the horizontal axis.

Against arterial PCO2, the human response is increasing and approximately linear over a broad range, with a slope that depends on the prevailing PO2, so a carbon dioxide curve needs its oxygen condition stated as well. Against alveolar PO2 it is strongly non-linear: shallow above about 60 mmHg and steepening as alveolar PO2 enters roughly 50 to 60 mmHg in Weil and colleagues’ normal subjects. Those subjects were studied under isocapnia, meaning carbon dioxide was added to the inspired gas to hold alveolar PCO2 at its resting value. Without that clamp, breathing harder lowers PCO2, hypocapnia and its alkalosis restrain ventilation both at the medullary chemoreceptors and at the carotid body, and the hypoxic response recorded is a drive and a brake that interact, not a simple sum. The curve itself changes too: in Weil and colleagues’ six subjects the fitted shape parameter differed significantly when alveolar PCO2 was held about 5 mmHg above or below control, so the shape depends on the prevailing PCO2.

So the resting controller in an adult at rest at sea level is a carbon dioxide controller. Arterial PCO2 is about 40 mmHg and arterial PO2 about 100 mmHg, which puts the operating point on the steep part of the carbon dioxide line and the flat part of the oxygen curve. A small PCO2 error produces a large ventilation change; a small PO2 error produces almost none. Oxygen is not irrelevant; its leverage is elsewhere on the axis. This lesson gives no universal central-peripheral split and no single carbon dioxide response slope, because reported values depend on the protocol used to get them.

Worked problem: what hyperventilation buys at altitude. These are stipulated model values, not measurements from any named mountain: barometric pressure 500 mmHg, inspired oxygen fraction 0.21, saturated water vapour pressure 47 mmHg, respiratory exchange ratio 0.80, negligible inspired carbon dioxide, steady state. The simplified alveolar gas equation is the tool, and ventilation and perfusion derives it:

alveolar PO2 = inspired humidified PO2 minus (arterial PCO2 / R)

1. Humidified inspired oxygen. 0.21 x (500 - 47) = 0.21 x 453 = 95.1 mmHg.
2. Before hyperventilation, with arterial PCO2 40 mmHg: 95.1 - 40/0.80 = 95.1 - 50.0 = 45.1 mmHg.
3. After hypoxic hyperventilation lowers arterial PCO2 to 30 mmHg: 95.1 - 30/0.80 = 95.1 - 37.5 = 57.6 mmHg.
4. Gain from washing out carbon dioxide. 57.6 - 45.1 = 12.5 mmHg. Independent check: (40 - 30)/0.80 = 12.5 mmHg.
5. Sea-level anchor, same assumptions: 0.21 x (760 - 47) = 149.7 mmHg, minus 50.0 = 99.7 mmHg, close to the alveolar PO2 of about 100 mmHg usually quoted for an adult at rest at sea level.

Three limits. It is alveolar PO2; arterial PO2 needs gas-exchange information the calculation does not contain. The 12.5 mmHg is the model’s output, not a measured acclimatization benefit. And the hyperventilation is held back at first by the alkalosis it creates. Bicarbonate loss partly corrects that alkalosis, while rising carotid body sensitivity and sensitization of the central pathways let ventilation climb further over days.

Barometric pressure PB 500 mmHg (760 is the lesson’s sea-level anchor) Arterial PCO₂ (PaCO₂) 40 mmHg (about 40 in an adult at rest at sea level)

Respiratory exchange ratio R

0.80 (the lesson’s value) 1.00 (hypothetical, not the lesson’s value)

Shade the alveolar PO₂ band of about 50 to 60 mmHg

Humidified inspired PO₂, 0.21 × (PB − 47)

95.1 mmHg

CO₂ term, PaCO₂ ÷ R

50.0 mmHg

Alveolar PO₂ (calculated)

45.1 mmHg

CO₂ washout gain, (40 − PaCO₂) ÷ R

0.0 mmHg

Calculated alveolar PO₂ against the isocapnic band of about 50 to 60 mmHg (one group of subjects; not a threshold; resting PCO₂ only)

below the band

In this sketch: steady state, a person breathing air, inspired oxygen fraction fixed at 0.21, water vapour pressure fixed at 47 mmHg and negligible inspired carbon dioxide; these are the lesson’s stipulated model values for its worked problem, not measurements from any named mountain. R = 0.80 is the lesson’s value; R = 1.00 is hypothetical, not the lesson’s value, and is invented for this sketch. The slider ranges are made up (760 mmHg is the lesson’s sea-level anchor), and the sketch does not say which combinations a person reaches. The reference PaCO₂ of 40 mmHg is the lesson’s value for an adult at rest at sea level. The band of about 50 to 60 mmHg comes from one group of normal human subjects, Weil and colleagues’, studied under isocapnia (alveolar PCO₂ held at its resting value); it is not a threshold. The values shown are alveolar, not arterial, PO₂. The sketch sets PCO₂ by hand and does not predict how much a person hyperventilates.

Alveolar PO₂ = 0.21 × (PB − 47) − PaCO₂ ÷ R. A lower barometric pressure lowers the first term; washing out carbon dioxide raises the result by (40 − PaCO₂) ÷ R. The rule gives alveolar PO₂, not arterial PO₂.

Altitude: the same loop, moved

At terrestrial altitude the inspired oxygen fraction stays at 20.94 percent; what falls is barometric pressure, and with it the partial pressure of oxygen in every breath. The same pathways now work from a lower input, their sensitivity changes over days, and their parts respond on different clocks.

Within seconds to minutes, arterial PO2 and saturation fall, carotid body firing rises and ventilation rises; saturations are read off the curve in carrying oxygen and carbon dioxide. Within minutes to hours, that hyperventilation lowers arterial PCO2 and raises blood pH, and the hypocapnic alkalosis inhibits ventilation both at the medullary chemoreceptors and at the carotid body, holding it below what the hypoxia alone would drive. Over days, renal compensation increases bicarbonate excretion and plasma bicarbonate falls, while carotid body sensitivity rises and the central translation of that input is sensitized, so ventilation can rise further even where cerebrospinal fluid pH stays alkaline; in Zouboules and colleagues’ lowlanders sampled by arterial puncture during a ten-day ascent to 5160 m, a derived index of renal compensation plateaued after about five days, a protocol-specific trajectory rather than a universal deadline. On a third clock, hypoxia elicits renal erythropoietin within hours, and erythropoietin raises red cell mass and oxygen-carrying capacity over longer exposure, as blood cells and hematocrit describes. Tibetan highlanders carry adaptively introgressed variation at EPAS1, the gene for HIF-2alpha, which human evolution covers.

Reasoning step: acute mountain sickness as a test of the loop. After rapid ascent in a non-acclimatized person every arrow above is in play at once, and the syndrome commonly presents as headache accompanied variably by appetite loss, nausea, fatigue or dizziness. A cerebral mechanism for uncomplicated cases is not settled. The 2018 Lake Louise revision, which removed disturbed sleep as a questionnaire item, is a research scoring instrument, and this lesson uses it only as that. What the loop does give you is directions: inspired PO2 down, arterial PO2 and saturation down, ventilation up, arterial PCO2 down, pH up, plasma bicarbonate down over days, erythropoietin up within hours, red cell mass up later.

Diving: the loop with the air supply cut off

Close the airway and the controller keeps running on a sealed store. Carbon dioxide accumulates and oxygen falls, and the apnea break point, the moment voluntary breath-holding ends, is influenced by both, and no simple blood-gas rule fully explains it; the relative weight of the two gases varies with the starting gases, the lung volume and the person. Overlaid on it is the diving response: in mammals, parasympathetically mediated bradycardia during apnea together with sympathetically mediated peripheral vasoconstriction. The autonomic limbs are the hypothalamus and autonomic system’s, the circuit evidence comes predominantly from terrestrial rodents, and the comparative diving-mammal story belongs to thermoregulation.

In people the response is studied by immersing the face in cold water during a timed apnea. Pernett and colleagues had 18 volunteers with little or no freediving experience perform two series of five maximal prone apneas with cold facial immersion and two-minute recovery intervals, one series preceded each time by 15 seconds of hyperventilation. Hyperventilation lowered mean starting end-tidal PCO2 from 29.0 to 17.4 mmHg (gas data usable in 13 of the 18), lengthened the mean apnea from 111 to 133 seconds, and lowered the mean saturation nadir from 93.6 to 90.6 percent, with no enhancement of bradycardia or vasoconstriction.

Read that as the loop predicts. Hyperventilation lowers the starting carbon dioxide and so delays the urge to breathe, while adding little oxygen to hemoglobin that is already nearly saturated. It does raise lung oxygen somewhat, and starting end-tidal PO2 in that study was higher after hyperventilation. The hold gets longer, the oxygen gets lower, and the diving response is no stronger for it. The mechanistic consequence discussed in that paper is that pre-dive hyperventilation may permit breath-holding to continue until extreme hypoxia produces blackout. Two limits: no blackout rate was measured in that protocol, and finger pulse oximetry reports peripheral hemoglobin saturation, not brain oxygenation. The ventilatory response to exercise is in exercise physiology.

Design problem

From the mouse result above, a colleague concludes that HIF-2alpha is the acute oxygen sensor of the glomus cell. Design the experiments that would separate a developmental loss of the cells from a failure of acute oxygen transduction in cells that are present.

One way to do it

Start from what makes the published design unable to answer it: a cell that is absent cannot report a sensing defect. The acute question needs an inducible or adult-onset deletion applied after the carotid body has formed, with glomus-cell number counted in the same animals tested physiologically. Without that count, a reduced response cannot separate missing cells from impaired sensing in the cells that remain.

Second, use a readout that does not depend on ventilation, because ventilation sums the whole reflex: glomus-cell calcium or catecholamine secretion to graded hypoxia in carotid body slices, or carotid sinus nerve afferent recording during a hypoxic step.

Third, keep the published controls: the carbon dioxide response as a control on the ventilatory apparatus, littermate controls, and the HIF-1alpha deletion as an isoform control.

Fourth, control the driver line. Tyrosine hydroxylase-Cre is active in adrenal chromaffin cells and brainstem catecholaminergic groups too, so a deficit could arise outside the carotid body; carotid body-targeted re-expression is the complement.

Then say in advance what each outcome means. If a verified adult-onset deletion spares the cells and the hypoxic response holds, that supports a mainly developmental role under the conditions tested, once deletion efficiency, timing and compensation have been checked. If normally numbered cells show an acute deficit, the sensing claim survives. Neither transfers to people without human evidence; the denervation data supply that for the organ’s role in breathing, not for HIF-2alpha.