BALTIMORE A sensory test pattern that jump-starts the link from the brain to the ear has been discovered to exist prior to the development of actual hearing capability. The discovery by Johns Hopkins University researchers here, once fully understood, may give electronic engineers an important new tool for building artificial cochleas and retinas.
This feedback control architecture long known but unexplained between the senses and the brain appears to be used by the body to craft the architecture of the sensor-to-brain neural network. That development appears to take place within the nervous system prior to the neural information processing experienced as hearing.
Johns Hopkins researchers found that the mammalian cochlea, or inner ear, can transduce acoustic signals to the brain through as many as 20 direct connections from the brain. After birth but before an animal hears outside sounds, it receives a modulated test signal from the brain's neural fibers, which it "hears" by echoing them back to the brain from the cochlea through a different network known as the afferent fibers. After this initialization stage, lasting several days, the animal appears to be ready to begin accepting inputs from the outside world, funneled in by the outer-ear canal.
"When rats are born, there is a membrane between the inner and outer ear, so that during the first few days of their life, the only signals they hear must come from the efferent feedback fibers coming from the brain to the inner ear," said Elizabeth Glowatzki, the principal researcher on the project, who co-authored the work with Johns Hopkins professor Paul Fuchs.
What they found in the laboratory was that before the animal's outer ear lets in any sounds from the outside world, the inner-ear hair cells are nevertheless spontaneously using the pathways between the ear and the brain. However, the researchers found overwhelming evidence from other wetware labs to support their theory that brains modulate this observed spontaneous signal by using efferent fibers.
"It has been known for a long time that efferent pathways existed between the senses and the brain, but nobody knew why," Glowatzki said. "We found that the brain gets signals from the ear's hair cells before the sensory cells can detect any signal from the outside world probably to help set up the neural connections between the senses and the cortex."
Prior to being exposed to the real world, the wiring between the senses and the brain is very dense almost "fully connected" in the sense that each nearby neuron is connected to every other nearby neuron. However, as the test signal from the brain is fed into the immature wiring matrix, most connections are pruned, leaving a sparsely connected, mostly feed-forward network, with limited local feedback paths.
"At first brains are wired with lots of connections to the senses, but during development many of these connections are retracted and the remaining ones grow larger and become well-established," said Glowatzki.
Pruning neural net
The specific laboratory experiment performed by the Johns Hopkins researchers was on a disconnected cochlea the inner ear, complete with hair-cell transducers so it did not show that the efferent connections from the brain provide the "training" input used to prune the neural network. But the team did show how the newly discovered spontaneous signal was used to prune the neural network prior to hearing, and the team cited literature to support the hypothesis that the efferent fibers are at work.
"It was known that there were connections from the brain to the hair cells, but nobody knew what they did," said Glowatzki. "Everybody thought it was some evolutionary leftover, but we have shown that these fibers are active before the animal hears anything, and our interpretation is that this feedback is used to help shape the connections to the brain before hearing can begin."
The Johns Hopkins researchers made their case by discovering the specific neurotransmitter used to activate the ear's hair cells before hearing begins. This neurotransmitter matched perfectly to the one already known to be released by the efferent fibers coming from the brain, making it the most likely candidate given current knowledge.
The next step for the team is to artificially stimulate the fibers, in lieu of the brain's feedback from the efferent fibers, and observe the result. Other cooperating labs are also planning to sever the efferent fibers at birth from test subjects and observe the effect that has on the animal's development.
"Next we will take out the inner-ear hair cells and put them in a culture, so that we can control all the signals and figure out what happens when the efferent fibers are missing. Others will record the signal from the hair cells to the brain before and after cutting the efferent fibers," said Glowatzki.
Developmental researchers studying morphology have already cut the efferent fibers in previous studies and observed a difference in the signal transmitted to the brain, but they had no theory to test and thus did not analyze the difference. Instead, the Johns Hopkins researchers armed with their theory that the signal jump-starts the connection matrix between ear and cortex should be able to perform an input/output analysis on the before-and-after recordings to deduce the transfer function of the feedback loop formed with the efferent fibers. "Eventually, we want to understand how this test signal from the brain helps to wire the sensory cells to the cortex," Glowatzki said. "But our next step will just be to verify that the brain uses the efferent fibers during the development of the pathways."