The Brain-Heart Axis: Circuitry From Insular Cortex to the Sinoatrial Node

01Insular chronotropic map

The insular cortex holds a chronotropic map, and its front and back move the heart in opposite directions

In 37 chloralose-anesthetized rats, tachycardia mapped to the rostral posterior insula and bradycardia to the caudal posterior insula Oppenheimer 1990. Phasic microstimulation locked to the ECG R wave produced heart rate changes with no accompanying change in blood pressure or respiration, so the cardiac effect was primary. Atenolol abolished both responses and atropine did not. The map moves the heart by raising and lowering sympathetic drive, and vagal outflow was not the route.

The same laboratory carried the finding into the human operating room. In five patients stimulated before temporal lobectomy, left insular stimulation produced bradycardia and depressor responses more often than tachycardia and pressor effects, at P below 0.005 Oppenheimer 1992. The right insula gave the converse pattern. Right-sided dominance for sympathetic effects is a human feature, and the rat is not lateralized this way. The stake named in that paper is sudden unexpected death in epilepsy.

Single-unit recording puts cells behind the bias. In six anesthetized monkeys, 81 neurons were recorded during cutaneous pinch and baroreceptor challenge, and twenty of the 26 units that answered blood pressure changes sat inside the insula, P below 0.001 Zhang 1999. Pressure-responsive units were 18 of 29 units recorded in the right insula against 2 of 14 on the left, P equal to 0.004. Two independent methods put cardiac sympathetic bias on the right.

Laterality is a bias held in a network

Both hemispheres produce both cardiac answers, and the axis that separates them runs front to back. Across 100 insular stimulations in 47 epileptic patients, bradycardia followed 26 stimulations and tachycardia followed 21, from either side Chouchou 2019. Tachycardia predominated posteriorly and bradycardia more anteriorly. Each answer carried its own spectral signature, with tachycardia raising the low- to high-frequency ratio and bradycardia raising high-frequency power.

Seizure data point the same way. In 13 consecutive video-EEG patients with ictal bradycardia, seizure activity was bilateral at bradycardia onset in nine Britton 2006. The authors state directly that their data do not support a strictly unilateral parasympathetic cardiomotor representation in the left hemisphere. Laterality in this axis is a bias inside a distributed network, and a cardiac effect reaches the heart through whichever hemisphere is available.

The Unified Model of Tone reads the split result as the expected one. An electrode delivers identical current on every trial, and the cardiac answer differs because the state receiving that current differs. A cortical site biases a system that already holds a rate.

02Findings

What the research shows

03Solitary nucleus and baroreflex

Cardiac afferents enter at the nucleus tractus solitarius, and the baroreflex leaves it through a GABAergic relay

Eighty percent of vagal fibers are afferent and 20 percent efferent Bonaz 2018. The nerve the body treats as an output cable carries four fifths of its traffic upward, into the nucleus tractus solitarius and the central autonomic network above it. Projections climb from there to the parabrachial nucleus and the periaqueductal gray, then on to the amygdala, the hypothalamus and the thalamus. That is where a heartbeat becomes information the rest of the nervous system can read and answer.

The solitary nucleus is not the only vagal entry port, and it does not handle viscera alone. Vagal afferents also terminate in the spinal trigeminal and paratrigeminal nuclei, and the solitary nucleus itself integrates somatic afferents alongside visceral ones Neuhuber 2022. Body-wall traffic and cardiac traffic meet at the first relay of the axis, before either reaches cortex.

Almost no barosensitive solitary neuron reaches the pressor region directly

Barosensitive neurons of the caudal ventrolateral medulla, not the solitary nucleus, carry sympathoinhibition to the rostral ventrolateral pressor cells. In rats made hypertensive for a sustained period, the activated caudal cells occupied a zone extending about 1.4 mm rostrally from the calamus scriptorius Chan 1998. Most carried GABAergic markers rather than glycinergic ones. Very few barosensitive solitary neurons projected straight to the rostral ventrolateral medulla. Baroreceptors in the aortic arch and carotid sinuses fire as vessel walls distend under rising pressure, and their afferents reach the pressor region only after this detour.

That relay sets how long sympathetic drive stays suppressed. Microinjecting the glycine receptor antagonist strychnine into the rostral ventrolateral medulla shortened baroreflex inhibition of renal sympathetic nerve activity in rats Gao 2019. It fell from 12 plus or minus 1 minutes to 5.1 plus or minus 1 minutes. In brainstem slices from the same laboratory, glycine carried less than 25 percent of the inhibition under steady state. With synaptic drive raised, the GABAergic share fell to 53 percent and the glycinergic share rose to 47 percent. Glycine carries little of the inhibition's magnitude at rest and much of its duration, which is why blocking it shortens the sympathoinhibition without weakening it. Baroreflex gain is not a fixed number, and its transmitter mix shifts with the state of the network.

The parasympathetic arm leaves the same nucleus. Solitary neurons excite the dorsal motor nucleus of the vagus and the nucleus ambiguus, raising vagal output while sympathetic drive is withdrawn, so both corrections are issued inside a single cardiac cycle.

04Vagal outflow to the heart

Vagal command of the heart runs in separate channels for rate, conduction and contraction

Parasympathetic command of the heart originates chiefly in the nucleus ambiguus, with a smaller contribution from the dorsal motor nucleus of the vagus. These efferents release acetylcholine onto cardiac ganglia and produce three negative effects. They are negative chronotropic and slow rate, negative dromotropic and slow conduction, and negative inotropic and weaken contraction. The three effects have separate addresses. Glutamate microinjected into the rostral ventrolateral nucleus ambiguus of the cat cut atrioventricular conduction rate by 76 percent and produced occasional second degree heart block, with heart rate unchanged Massari 1995. Ipsilateral cervical vagotomy abolished the effect.

That study also counts the anatomy. The dorsal motor nucleus held one third as many labeled cells as the ventrolateral nucleus ambiguus. The left rostral ventrolateral nucleus ambiguus held three times as many labeled cells as the right, P below 0.025. Left vagal dominance over atrioventricular conduction has a cell count behind it.

Ultrastructural work in the cat maps the three channels to three ganglia. Separate populations of ventrolateral nucleus ambiguus preganglionic neurons project to the sinoatrial ganglion for rate, the atrioventricular ganglion for conduction and the cranioventricular ganglion for left ventricular contractility Blinder 2005. Enkephalinergic terminals occupied 8 percent of axodendritic synapses on negative chronotropic neurons and 12 percent on negative dromotropic ones. The vagal brake is three addressed lines, and each line is separately modulated.

A tonic vagal layer sits beneath the reflex

A steady vagal influence on the heart runs underneath the beat-to-beat baroreflex, independent of breathing and of the cardiac cycle. Extracellular recording in the dorsal vagal motor nucleus of anesthetized rats found 56 neurons with non-myelinated axons Jones 1998. Sixteen of these C fiber cells carried ongoing activity with no relationship to central respiratory drive, lung inflation or the cardiac cycle. Twenty-one of the 56 fired at short latency once phenylbiguanide was injected into the right atrium, so the input they answer is the pulmonary C fiber. A separate sample of 33 cardiac vagal preganglionic neurons recorded in anesthetized cats behaved the same way, with no respiratory rhythm and no sign of arterial baroreceptor input.

What high-frequency variability actually reads

High-frequency heart rate variability from 0.12 to 0.40 Hz tracks vagal influence on the heart and not sympathetic influence. Blockade settled it: in women studied under atropine and metoprolol, the high-frequency band moved with vagal influence alone, while the preejection period did the reverse and moved with sympathetic influence alone Cacioppo 1994. No other common measure gave a specific index of tonic activation in either branch.

The index is not a straight scale. Eight healthy young men received nine consecutive intravenous atropine doses, with seven minutes of ECG and respiration recorded at each dose, and the measurements were fitted to a model of vagal control Pyetan 2003. Mean heart rate fell along a curve whose fitted parameters barely varied between subjects. Respiratory sinus arrhythmia did not: the parameters of its fitted curve, and the shape those parameters produce, account for the conflicting published results on variability as a vagal index. High-frequency power reports where vagal control sits on a subject-specific curve, so equal changes in the index do not mean equal changes in vagal traffic.

05Hypothalamic and prefrontal control

The hypothalamus and medial prefrontal cortex set the gain of the brainstem cardiac reflexes

The lateral hypothalamic area holds an obligatory relay on the route from insula to heart. In 28 chloralose-anesthetized rats, microinfusing 390 nl of the glutamate antagonist kynurenic acid into the lateral hypothalamic area attenuated insular tachycardia by 95 percent, and the synaptic blocker cobaltous chloride abolished it Oppenheimer 1992. Naloxone attenuated it by 95 percent as well. Neuromodulators scaled the route in the other direction. Leu-enkephalin and neuropeptide Y doubled the cardiac response, and dynorphin, a kappa opioid agonist, augmented it eightfold. A cortical command reaches the heart through a synapse whose strength is set by peptides.

The cortex resets a brainstem reflex through the solitary nucleus

Medial prefrontal cortex sends projections down to the solitary nucleus, and silencing that nucleus removes most of the cortical effect. Bilateral muscimol, 44 pmol per 25 nl, silenced the rat solitary nucleus. That removed 62 percent of the prefrontal depressor response and 65 percent of the sympathoinhibitory response Owens 1999. The cortico-solitary projection is excitatory, and it works by recruiting the intramedullary baroreflex pathway rather than bypassing it.

Gain-setting and baseline-setting are separable operations at that nucleus. In rats, stimulating infralimbic prefrontal cortex facilitated aortic baroreflex bradycardia by 65 percent electrically and 60 percent chemically, while preoptic hypothalamic stimulation facilitated it by 70 percent and 69 percent Sévoz-Couche 2006. Blocking 5-HT2A receptors inside the solitary nucleus prevented all of that facilitation and left the baseline pressure and rate changes intact. One receptor population at one nucleus carries the reflex gain and leaves the resting value alone.

One cortical site engages more than one target. Prefrontal stimulation raised Fos-like immunoreactivity in the rostral, intermediate and caudal ventrolateral medulla and in the dorsal motor nucleus of the vagus as well as in the solitary nucleus Owens 1999. So cortex holds two handles on one beat. It can raise the vagal answer to pressure through 5-HT2A receptors at the solitary nucleus, and it can reach the pressor region that sets sympathetic drive.

06Neurogenic cardiac injury

Central nervous system events blunt cardiac contraction with open coronary arteries

Neurogenic stress cardiomyopathy appeared in 16 percent of patients with aneurysmal subarachnoid hemorrhage, defined by a depressed ejection fraction plus a regional wall motion abnormality in a non-vascular pattern Kilbourn 2013. Mortality reached 46.9 percent in that group against 11.2 percent in the rest, P below 0.001. The non-vascular distribution is the tell, because coronary ischemia damages the territory its vessel feeds. Contraction failed here in a pattern no coronary anatomy can draw.

A registry of unselected cases puts numbers on the wider syndrome. Across 1,750 patients in a 26-center international takotsubo registry, mean left ventricular ejection fraction was 40.7 percent against 51.5 percent in age- and sex-matched acute coronary syndrome patients, P below 0.001 Templin 2015. Neurologic or psychiatric disorders were carried by 55.8 percent of the takotsubo patients against 25.7 percent of the comparison group. Emotional triggers are proven and account for 27.7 percent of cases, physical triggers for 36.0 percent, and 28.5 percent of patients had no evident trigger. That is why the load-bearing finding in this registry is the neurologic and psychiatric history and not the trigger. The condition is not benign, running 9.9 percent major adverse cardiac and cerebrovascular events and 5.6 percent death per patient-year.

The lesion data name the cortex

In 179 acute stroke patients, a right insular lesion raised the odds of death within three months more than sixfold Christensen 2005. The odds ratio was 6.2, with a 95 percent confidence interval of 1.5 to 25.2, and it held independent of stroke severity, lesion volume and age. Insular lesions were present in 43 of the 179 on CT, 25 left-sided, 17 right-sided and one bilateral. They tracked with sinus tachycardia above 120 beats per minute at P equal to 0.001. They also tracked with ectopic beats above 10 percent of recorded beats at P equal to 0.032, and with ST elevation at P equal to 0.011. Right insular lesions specifically tracked with atrial fibrillation, atrioventricular block and inverted T waves. Those are deflections of the PQRST waveform, the five waves that make up one beat on the surface electrocardiogram.

Continuous recording shows the same lesion collapsing the beat-to-beat range. In 103 consecutive first-ever stroke patients on 24-hour Holter, right insular damage lowered SDNN and rMSSD and raised the low-frequency to high-frequency ratio, at P below 0.05 Colivicchi 2004. Those patients also carried more complex arrhythmias than any other lesion location. The heart lost its range before it lost its rhythm. A heart with pristine coronary arteries can fail electrically because a few cubic centimeters of cortex have gone dark.

07Stellate ganglion remodeling

After infarction the cardiac sympathetic supply regrows unevenly, and the stellate ganglion changes with it

Myocardial infarction injures the nerves inside the heart before it changes the nerves outside it. Intrinsic cardiac denervation is followed by sympathetic nerve sprouting at the peri-infarct zone, and all three levels of the cardiac neuraxis remodel: cortical, medullary and intrinsic. Driving that sprouting shows what it costs. Dogs with chronic infarction and complete atrioventricular block received nerve growth factor at the left stellate ganglion, and spontaneous ventricular tachycardia ran tenfold higher than in dogs without it Cao 2000. Phase 2 episodes per day were 2.0 plus or minus 2.0 in the nerve growth factor group against 0.2 plus or minus 0.2 in controls, at P below 0.05. The timeline came in two phases: early tachycardia persisted 5.8 plus or minus 2.0 days, then a second phase reappeared 13.1 plus or minus 6.0 days after surgery.

The endpoint followed the sprouting. Sudden cardiac death occurred in four of nine dogs given nerve growth factor after infarction and atrioventricular block, and in none of six dogs without it Chen 2001. In explanted native hearts from human transplant recipients, nerve density correlated positively with a clinical history of ventricular arrhythmia. Regrowth is dangerous because it is uneven. Heterogeneous reinnervation produces regional differences in repolarization, and that dispersion is the substrate for ventricular tachycardia and fibrillation.

Inflammation inside the ganglion is enough on its own

Interleukin-1 beta was injected into the left stellate ganglion of 24 canines, in three groups of eight Wang 2017. It shortened the effective refractory period and APD90, steepened the maximal slope of the restitution curve and increased ventricular arrhythmia. Nerve growth factor and neuropeptide Y rose inside the ganglion while neuronal nitric oxide synthase fell. Pre-injecting the interleukin-1 receptor antagonist attenuated all of it. One cytokine remodeled a sympathetic relay, with no infarct required, and the heart inherited the change as arrhythmia.

The glia in that ganglion set the gain. Two hours of myocardial infarction in rats provoked satellite glial cell activation, and the activation correlated positively with cardiac norepinephrine release Zhou 2025. Chemogenetic silencing of those glia cut sympathetic hyperexcitability, stabilized ventricular electrophysiology and reduced nerve sprouting at 7 days. P2Y1 receptor and IGFBP2 signaling carried the glia-to-neuron message. The handle is state-dependent. Exciting those glia destabilized the ventricle in uninjured rats, while silencing them changed nothing in a heart that had not been infarcted.

08Somatic input to the heart

Somatic input reaches the brainstem neurons that drive cardiac sympathetic outflow

Median-nerve stimulation in anesthetized rats inhibited the splanchnic-evoked responses of presympathetic rostral ventrolateral medulla neurons, recorded from the cells themselves Zhou 2005. Electroacupuncture cut those responses by 49 plus or minus 6 percent and manual acupuncture by 46 plus or minus 4 percent, both delivered at about 2 Hz. The inhibition lasted 20 minutes after the stimulus ended. All 18 neurons recorded received baroreceptor input, and 12 were confirmed presympathetic by antidromic drive from the intermediolateral columns at T2 to T4, the segments that supply the heart.

Frequency scaled the answer these cells gave. Electroacupuncture at 2 Hz evoked 53 plus or minus 10 impulses per 30 stimulations in these neurons, and at 100 Hz it evoked none at all. The fall between them was graded, running 16 impulses at 10 Hz, 8 at 20 Hz and 2 at 40 Hz. Current was held between 0.3 and 0.5 milliamps at every frequency, so intensity is not what moved these cells. One nerve, one population of neurons and one anatomy gave everything or nothing depending on the rate at which information arrived.

Somatic and cardiac traffic also converge in cortex

In monkey insula, 16 of the 20 blood-pressure-responsive units also fired to noxious pinch Zhang 1999. Their receptive fields were wide and bilateral, covering face, hand, foot and tail. At the other end of the axis, the solitary nucleus integrates somatic afferents alongside visceral ones Neuhuber 2022. Body-wall information and cardiac information share cells at the first relay and share cells again in cortex.

Chiropractic care loads thoracic joints and the receptors around them at T2 to T4, the levels from which these presympathetic neurons were antidromically driven. In the Unified Model of Tone, the cardiac consequence of a somatic input is set by the segment it enters and the rate at which it arrives, and force sits on a separate axis. The graded 2 Hz to 100 Hz curve is where the rate half of that claim can be checked. At one nerve, in one population of medullary cells, 2 Hz cut visceral-evoked firing by about half and 100 Hz did nothing.

Tone inside its healthy range shows up in this axis as a heart free to move rate, conduction and contraction where the moment demands and return afterward. Tone driven outside that range shows up as a rate that will not fall, a beat-to-beat range that has collapsed and repolarization that has grown uneven across the wall. Each of those three has a measured endpoint: sinus tachycardia above 120 beats per minute, a fallen SDNN and the dispersion that follows uneven sprouting.

09Tone

How this system expresses tone

Tone is the organization the nervous system holds, and the brain-heart axis is the one place it can be read in milliseconds and millivolts. Coupling, gain and input quality do the most work here.

Each remaining foundation has a cardiac reading. Set point: cardiac vagal neurons hold a resting bias, and 16 of 56 dorsal vagal C-fiber neurons in rats fired steadily with no tie to breathing or to the cardiac cycle. Oscillation: vagal traffic arrives in bursts timed to respiration, which is why the 0.12 to 0.40 Hz band tracks vagal influence and the slower bands do not. Prediction: cortical sites move the heart before any metabolic demand arrives, which is how 500 microamps at an insular site changes rate in an anesthetized animal lying still. Time course: baroreflex sympathoinhibition lasted 12 minutes in control rats and 5.1 minutes once glycine receptors in the rostral ventrolateral medulla were blocked. Load: sustained catecholamine drive is paid for in muscle, with ejection fraction falling to 40.7 percent across 1,750 stress cardiomyopathy patients. Constraint: the vagal hold on conduction is limited by the column that carries it, with three times as many labeled cells in the left rostral ventrolateral nucleus ambiguus as in the right.

10Across the library

How this page relates to the rest of the library

11Frequently asked

Questions about this topic

12The sources

References

Sources: primary literature, linked inline.