=Amphibian Papilla

Ellen Leverenz

In the first phase of our AP dye tracing studies, Ellen, Hironori and I mapped twenty three afferent axons whose frequency sensitivity had been determined with open-field stimuli. Thirteen of those axons were traced all the way to their peripheral terminations- on hair cells. Ten were traced far enough to determine which small region of the AP each of them innervated, but not all the way to their target hair cells. Those twenty three fibers demonstrated that the AP was polarized with respect to frequency response. All of the fibers with low best excitatory frequencies (low BEFs) innervated hair cells exclusively in the rostral patch. All of the fibers with high BEFs innervated hair cells in the caudal extension. Those with intermediate BEFs innervated hair cells in the vicinity of the tectorial curtain.

In the second phase, Ellen and I traveled to the Capranica Lab at Cornell, where we spent three weeks working with Bob Capranica, Andrea Megela Simmons and Gary Harned. There we traced our first fibers characterized with closed-field stimuli. Seven of these were traced to the AP. On our return to Berkeley, Ellen, Hironori and I used the approach we had learned at Cornell to characterize and trace approximately ten more AP afferent axons. At that point, we had enough data for our paper (The tonotopic organization of the bullfrog amphibian papilla, an auditory organ lacking a basilar membrane, J. Comp. Physiol., 145: 437-445 (1982)).

In the process of mapping frequency sensitivity in these units, we had (tape-) recorded the stimulus and response signals for sixteen. Ellen subsequently added six more units and set out to analyze the data for properties other than BEF, properties such as threshold, phase locking, and degree of adaptation. Working in Bob’s lab, Andrea had pioneered studies of adaptation in the bullfrog AP, and collaborated with Ellen on that part of the project. Altogether, they analyzed twenty-two units, two from open-field studies, twenty from closed-field studies. The locations of the peripheral origins of these units are shown in the Tonotopy map.

The results of the adaptation, threshold (sensitivity), and phase-locking analyses were published in Ellen’s dissertation, but not elsewhere. Twenty five years later, all of these results remain timely. This, I believe, is a consequence of the shift away from VIIIth-nerve physiology as popular endeavor. Current studies of the vertebrate ear tend to be molecular or micromechanical-- deeply reductionist.

For many years, beginning perhaps in the 1960s, the life sciences have diverged into two huge and productive areas-- integrative biology and cellular & molecular biology. As this unfolded, I was struck by the movements away from traditional systems-physiological and morphological studies. It seemed to me that these occupied the middle ground-- between the cellular/molecular and the integrative. It also seemed to me that, as the deeply reductionist side of biology turned to face the complexity of life as it really is (in an empirical sense, not an ontological sense), that middle ground might prove useful. On the other hand, the generation of the new Amphibian Tree of Life is demonstration of how easily the gap between the molecular and the integrative can be spanned directly by ingenious researchers. When I first arrived at Berkeley in the summer of 1967, I believed that the history of life on Earth would be forever veiled by the mist of time, that strong inference was out of reach in evolutionary studies. Shortly after I arrived, I heard a talk by Vincent Sarich, about his work with Allan Wilson. The veil was lifted, strong inference was achievable. It reminds me of those lines by Alexander Pope -- ". . . . and all was light." I was glad to be at Berkeley, among the shoulders there that might allow me to see a little farther.

Regarding Ellen's analyses, I believe we rationalized her phase-locking results with subsequent Wiener kernel analyses, which I shall address below. In anticipation of reuniting with Ellen and Andrea to help put their adaptation data and results into the archival literature, I present here just a very brief summary taken from Ellen’s dissertation. From 24 items, I include five of the six that are related to morphology-- i.e., to location on the AP or to number of hair cells innervated.

From E.L. Leverenz,, Signal Processing in the Amphibian Papilla of the Bullfrog , University of California, Berkeley, doctoral dissertation (Biophysics), 1988.

“These are the principal conclusions that can be drawn from this investigation:

(1) The amphibian papilla is tonotopically organized along its rostro-caudal axis, with the higher-frequencies at the caudal end of the papilla. New data do not support an earlier hypothesis of tonotopic organization within the “head” region [caudal patch] of the AP, but some evidence suggests a tonotopic gradient across the papilla (medio-lateral, with higher frequencies on the lateral side).

(2) Afferent nerve fibers innervate from one to ten or more hair cells. Fibers innervating the rostral (low-frequency) region of the papilla tend to contact more hair cells than those innervating the central and caudal regions, but no region is exclusively innervated by fibers contacting few or many hair cells.

(4) The number of hair cells innervated by a fiber does not seem to correlate with the fiber’s sensitivity or sharpness of tuning.

(5) Fibers that innervate type A [juvenile] hair cells have low or moderate sensitivity. Other hair cell types do not seem to correlate with sensitivity level.

(19) Weakly and moderately adapting fibers innervate many hair cells, and strongly adapting fibers innervate only one or a few hair cells.”

The sixth item (Ellen’s number 14) is related to nonlinear phase-locking and is discussed under “Nonlinear properties of the bullfrog AP.”

In addition to coauthoring several papers while she was a member of the Lewis Lab, Ellen coauthored a monograph The Vertebrate Inner Ear (CRC Press, 1985). Her timeless 82-page review of “Comparative inner ear anatomy” in that monograph continues to be cited.

Kapu

Using stimuli comprising just one sinusoidal frequency at a time, Ellen did not look, explicitly, for two-tone suppression. She did not address the question of whether or not the non-suppressible units were strictly confined to the caudal extension. Unfortunately, we had not yet discovered the power of the second-order Wiener kernel to reveal suppression. If we had, we would have used noise stimuli instead of sinusoids—and we would now have the answer to that question.

Among the twenty-two AP units that Ellen and Andrea analyzed, only one exhibited significant background spike activity in the absence of investigator-applied stimuli. Studying that unit thoroughly, Ellen found that the background spike activity was suppressed by sinusoidal stimuli at frequencies slightly below the unit’s BEF and by sinusoidal stimuli in a range above the unit’s BEF. She described these results carefully in her dissertation.

A colleague, who shall be un-named here, had seen the same thing (in the bullfrog) during his doctoral research, but had been told by his mentor that he should ignore it - - it couldn’t happen. He saw it again (in a bird) as a postdoc, and again was told, by another mentor, to ignore it. Among a subset of auditory physiologists, such an observation (suppression of background spike activity) was a sure sign of sloppy technique. “If the activity is suppressible, it must be stimulus-driven; and the culprit stimulus must be leaking into the experimental system.”

In the same year that Ellen published her dissertation, Ken Henry and I made the first presentation of our work on afferent nerve fibers from the gerbil cochlea. Quite incidental to the points we were making, some of our histograms showed suppression of background spike activity. We were well aware that this was kapu, but we also were aware that this affect already had been reported for gerbil cochlear axons, and that we had seen it clearly in published histograms for other mammals. Nonetheless, the evidence of suppression in our histograms became the central theme of the discussion period following our presentation. Our apparatus surely was flawed. Later that day, a co-author of one of the “other mammal” papers denied that he’d ever seen suppression of background activity.

We decided to carry out a set of experiments that would show conclusively that the suppressible background spike activity in the gerbil was not the result of leaks - - not through our amplifier system, not through the walls of our acoustic isolation chamber. Basically, we were reaffirming what already was known to those immune to the kapu, in a subset of gerbil cochlear units, background spike activity (not driven by stimuli from outside the animal)is suppressible. In a companion paper, we showed that mid-frequency units from the gerbil cochlea produced spikes in response to the effects of the animal's heartbeat-- as one would expect. We expect gastric sounds and respiratory sounds also lead to spikes in cochlear units. The BEFs of the units in which we observed suppression were well above those of units that responded to heartbeats, and the background spike activity was not correlated with respiration. We suspect that it arose from Brownian motion in the inner-ear micromechanical milieu, in the noisiness of channel activity in the unit's haircells, or in noisiness in its synapses.

What was it that Bacon said—“If you want to understand nature, look to nature itself, not to books [not even to mentors].” This kapu clearly belongs with Bacon’s “idols of the theatre.”

Xiaolong Yu

Xiaolong's work on dithering in the Hodgkin-Huxley model and on the bullfrog sacculus is discussed in the Sacculus section. Here I show only his conclusions regarding adaptation and suppression in the bullfrog amphibian papilla. He was among the first members of the Lewis lab to employ de Boer's REVCOR tool. REVCOR combines white noise stimulus and triggered correlation to yield an estimate of the filter function (impulse response)of the peripheral tuning structure associated with each inner-ear unit. (Recall that inner-ear unit comprises the afferent axon, the hairs cells connected to it, and the mechanical structures linked to those haircells). This gave him a powerful advantage over Ellen, who had used stimuli (trapezoidally modulated sine waves) that were more standard in hearing research at the time. Xiaolong used band-limited white noise stimulus not only for REVCOR analysis, but also as background when he applied tone bursts (trapezoidally-modulated sine waves).

Regarding adaptation and suppression in the bullfrog AP, here are some conclusions from his research.

From Xiaolong Yu, Signal Processing Mechanics in Bullfrog Ear, Inferred from Neural Spike Trains, University of California, Berkeley, doctoral dissertation (Electrical Engineering and Computer Science), 1991.

Conclusions regarding adaptation:

Tuning changes

The impulse response of the tuning structure of each AP unit changes systematically with increasing background noise level (short-term root-mean-square noise amplitude). As the level increases, the impulse response becomes shorter (more highly damped).

Following a stepwise change in noise level, the change in the impulse response takes place in a time that is short relative to the duration of the impulse response itself.

Gain and bias changes

Adaptation to a tone burst at or near BEF comprises two components-- a decrease in ac gain and a negative shift in dc bias (see "Sacculus physiology from the outside, part II").

In the presence of continuous background noise stimulus at a constant root-mean-square amplitude, the magnitudes of these changes increase as the amplitude of the tone burst increases relative to that of the noise.

Adjustments of ac gain and dc bias tend to keep the positive peaks of the stimulus waveform (background noise plus tone burst) very close to threshold. Adaptation, then, is simply the transient associated with maintenance of this situation.

Suppression

Adaptation to a strong tone burst at BEF can suppress a unit’s ongoing response to broad-band noise. A tone burst at frequencies somewhat above the unit’s tuning curve will do the same thing, but the responses in the two cases are qualitatively different. At BEF, the suppression (adaptation) comes on gradually, and it recedes gradually when the tone burst ends. At the higher frequencies, the suppression comes on abruptly; and it ends abruptly (frequently with overshoot) when the tone burst ends. Overshoot does not occur with BEF-tone suppression (adaptation).

While adaptation appears to involve processes within the unit itself, suppression by the higher-frequency tone bursts seems to involve a blocking (or partial blocking) of the stimulus before it reaches the unit.

Walter Yamada and Greg Wolodkin

Realizing that using first-order Wiener kernels limited our analysis to low-frequency units (units with BEFs below approximately 500 Hz) I decided that we should try second-order kernels. Pim van Dijk already had applied these successfully to the AP of a European frog; and after a visit to his lab, I asked Walter to do the same thing for us. I wanted to develop strong inferences regarding the dynamic order of the higher-frequency auditory tuning structures (the number of integration processes between the point of stimulus input and our observed responses). This would tell us much about the underlying biophysical structures as well as much about signal-processing strategies sculpted for the frog's ear by evolution (see “Signal Processing in the Frog’s Ear”). Furthermore, it was not an issue of concern to Pim, so we would not be competing with him. Walter used Pim’s algorithm and quickly began generating second-order kernels for the bullfrog AP.

By this time, Walter had been joined by Greg, and the two of them soon faced a puzzle. Second-order kernels for lower-frequency units exhibit a checkerboard pattern of alternating positive and negative values; those for higher-frequency units exhibit a pattern of parallel, alternating positive and negative lines. The checkerboards are easily explained; they are formed by the sum of the outer products of the corresponding first-order kernels (see Saccular Physiology from the outside). The parallel lines were not easily explained. Walter and Greg decided to decompose the second-order kernel into constituent vectors (by singular-value decomposition) in an attempt to learn how the parallel lines could arise. We soon realized that the lines could be produced by the sums of the outer product of two vectors, 90 degree out of phase with one another (i.e., in quadrature with one another). Applying the formulation of the second-order term of the Wiener Series to these quadrature pairs of vectors, we realized, further, that they bear considerable significance regarding the nonlinear filter properties of the AP (again, see Physiology of the sacculus from the outside).

Walter found that in the decomposed kernels he could separate nonlinear excitatory effects from nonlinear inhibitory effects. From that it was a simple step to reconstructing an inhibitory sub-kernel and an excitatory sub-kernel, the sum of which was the original, full second- order kernel. Interpretation of second-order Wiener or Volterrra kernels has long been a problem; and I believe this was the first time one had been decomposed into separate excitatory and inhibitory sub-kernels. Each of these was easily interpretable separately. The kernel was beginning to yield to a new interpretive method. (Pim already had successfully applied another interpretive method—translation to a sandwich model.)

These revelations revolutionized the Lewis Lab’s approach to auditory physiology, and with data taken many years ago— by former students now well into their professional careers, I’m still enjoying their fruits.