# Circuits: Amplification, Feedback, Adaptation and Switches

> Circuit architecture and kinetics shape a signal's size, duration, and reversibility. Follow one photon through a rod's amplifying cascade, then compare feedback circuits that limit a response, make a switch, or return an output toward rest. Afterwards you can compute a simple cascade's gain, explain the costs of amplification, distinguish ultrasensitivity from memory, and explain robust adaptation in bacterial chemotaxis.

Course: Cell Signaling (How a cell knows anything). Level: advanced university and graduate. About 20 minutes.
Lesson page (link this one when citing): https://stemhop.com/lessons/signaling-circuits/

Builds on: Binding: Affinity, Occupancy and Dose–Response · A Field Guide to Receptors

## What you'll learn

- Use circuit words (gain, negative and positive feedback, feed-forward, adaptation, switch-like response, robustness, insulation, cross-talk) to describe a pathway.
- Compute the gain of a cascade as a product of its steps and state what amplification costs.
- Explain how delayed negative feedback can set how long a signal lasts and how sufficiently strong positive feedback around an ultrasensitive step can make a decision that does not reverse.
- Explain adaptation using one real example.

One photon, absorbed by one rhodopsin molecule in a rod cell, changes the current through that cell by about one picoamp, a trillionth of an ampere. A single molecule cannot carry that current. How does its signal become large enough to register, yet remain local and stop?

**The one idea.**

Circuit architecture and kinetics shape a signal’s size, duration, and reversibility.

**Scope:**

Use this rule for a defined pathway, cell and input. Wiring alone is not enough: component abundance, catalytic rate, active lifetime, saturation, localization and delays matter. Step gains multiply in the simple unsaturated cascades used here, where each active molecule independently reaches a fixed number downstream.

A pathway diagram is only a wiring list. To predict behaviour, ask how much output one input produces, how output changes with time, and whether the response can reverse.

## Gain is a product, within a simple model

The first film turned one photon’s path through rhodopsin, transducin and phosphodiesterase (PDE) into a product of step gains. Here are the measured numbers.

In darkness a rod carries a **dark current** of about 20 picoamps through channels that cGMP holds open. Denis Baylor, Trevor Lamb and King-Wai Yau recorded the single-photon dip from one toad rod in 1979: about one picoamp, a few percent of that current.

*Figure (AI-generated illustration).* One photon activates a multiplying relay that lowers cGMP, closes some channels and reduces the dark current. Image: a chain of rhodopsin hit by a photon, then transducin, then PDE, then falling cGMP; an inset compares a membrane in the dark with channels open to one after the photon with some of them closed.

**Gain** is output molecules per input molecule. Rhodopsin to transducin and PDE to cGMP are **catalytic** steps, where one active molecule changes many downstream molecules while it stays active; the transducin-to-PDE handoff is **one-to-one** and multiplies nothing.

Leskov and colleagues measured the first step in isolated rod membranes as a rate: one active rhodopsin switches on transducin, and through it PDE, at 120 to 150 molecules per second. Rhodopsin is switched off quickly, though, so in an intact rod one photon activates only about 16 transducins in mouse and about 60 in frog. Those PDE molecules then destroy about 2,000 cGMP molecules in mouse and about 72,000 in frog.

**Predict.** Use the mouse numbers. Does more of the gain lie in the first catalytic step or the last?

> **Check your prediction.** The last. The first step multiplies one rhodopsin into about 16 transducins. The last multiplies those 16 into about 2,000 destroyed cGMP molecules, roughly 125 per active PDE. In frog the split is about 60 against about 1,200.

The interactive uses a different cascade, epinephrine to glycogen breakdown. Change one step at a time.

*Interactive on the lesson page:* Cascade amplifierset gains · flip switches Epinephrine molecules bound Knock out one step: — glucose molecules mobilized

## Amplification sends a bill

The first film named three costs: noise, shutoff and spread. The rod shows how the first two are paid; other pathways show the third.

First, noise: amplification can magnify upstream fluctuations, and a one-picoamp dip sits on a baseline that never stops moving. Baylor’s group could trust the dip only because they counted. They dimmed flashes until most trials were blanks; if each response is one photon caught at random, the blank fraction must follow a Poisson distribution, the signature of rare independent events. It did.

Second, shutoff. Rhodopsin kinase and arrestin shut down rhodopsin. The RGS9 complex speeds transducin’s self-shutoff by GTP splitting. The rod also uses **negative feedback**, output acting to reduce its own production: closing channels lowers calcium inside the cell, and calcium-sensing GCAP proteins speed the enzyme that makes cGMP.

Third, a diffusible messenger made in many copies risks reaching unintended targets. A cell that answers different receptors differently must keep each pool of messenger near where it was made. mAKAP, for example, holds a kinase, a messenger-destroying phosphodiesterase and the kinase’s target within nanometres, so the cyclic AMP that matters stays in that small neighbourhood. That is **insulation**, keeping one pathway’s signal away from another pathway’s targets, and it has to be built.

Amplification creates output; shutoff and feedback loops, like the rod’s calcium loop, decide how that output changes with time. The second film compares three feedback designs under a continuing input.

**Film transcript: Three Ways a Signal Changes with Time**

### Film transcript: Three Ways a Signal Changes with Time

The first film showed a cascade multiplying one input, then controlling its spread and shutoff. What can make an output fade, commit, or reset while input remains? One answer lies in how output loops back. Feedback begins when output changes its own production. A return arrow that reduces production is negative feedback. A return arrow that increases production is positive feedback. Timing and strength decide the result. The same pathway can therefore produce different dynamics.

Connections alone cannot predict the time course. Start with a continuing input and a delayed brake. Output rises before the brake accumulates. Then negative feedback lowers pathway activity, even though input remains. The brake's delay and strength help shape the pulse. Connections alone did not predict it. Now turn the return arrow forward.

Output strengthens its own production, so a small rise can recruit more activity. If the loop is strong and includes a steep step, the circuit can settle into low or high output. A weaker loop may only sharpen the response. An ultrasensitive response is steep, but steepness alone can reverse when input falls. Sufficient positive feedback can add bistability and memory. Then a temporary input may leave output in the high state. The memory outlasts the input that set it. A third circuit receives a continuing input. Three output traces appear on screen. Which trace represents adaptation?

Look for the defining pattern before choosing. Adaptation is the return toward rest while stimulus remains. So choose the trace returning toward its starting line. A delayed brake can make a similar shape. Integral feedback is a special negative feedback. It keeps correcting the gap from a reference output until the gap is zero. In bacterial chemotaxis, receptor methylation acts as a running total of that gap in an idealized model.

Correction continues until the error reaches zero. Engineers call this integral feedback. CheR adds methyl groups, while CheB removes them. Changing enzyme amounts can change how quickly correction proceeds and can shift other steady behaviour. Yet adaptation precision can remain robust. Robustness protects a specified property, not every feature of the response. Place the three designs side by side. Delayed negative feedback can limit duration. Sufficient positive feedback can sharpen commitment and preserve memory.

Integral feedback corrects until the error is zero. It can restore a reference output despite changing parts. The pathway's arrows begin the explanation, but kinetics finish it. Loop sign, strength, delay, and component lifetime determine whether output fades, commits, or adapts. Feedback reshapes time and reversibility.

From the film

- Feedback can limit duration, sharpen decisions, or restore resting output.
- Delayed negative feedback can make output fall while an input remains present.
- Sufficiently strong positive feedback can add bistability and memory to an ultrasensitive response.
- Integral feedback, a negative feedback that corrects until the error is zero, can make adaptation precise even when protein amounts change.

## Negative feedback can set how long a signal lasts

The second film drew a delayed brake; here is a molecular design for one.

**Predict.** You have a growth factor, its receptor, a kinase and one protein X. How could X make kinase activity fall while the growth factor remains present? What must X sense, what must it act on, and what timing must it have?

> **One way to do it.** Make X a phosphatase whose amount rises when the kinase is active and which removes the kinase’s phosphates.
>
> - X senses the kinase’s output, so it appears only after the pathway has fired, and it acts on the kinase, so activity falls while the growth factor is still there.
> - X acts late. If it worked as fast as the kinase, the signal would barely rise. If it must be made from scratch, the delay is minutes, and that delay sets the window.
> - X is short-lived. If it lingered once the growth factor was gone, the cell could respond only weakly a second time.
>
> Cells use this design: ERK drives production of DUSP-family phosphatases that remove ERK’s own phosphates.

The brake’s delay, strength and lifetime help set the activity window, but this is one design, not a universal timer: receptor traffic, component degradation and other loops can also set duration.

## Positive feedback turns a dial into a switch

The second film drew an idealized switch. The experimental question is whether a smooth population average can hide individual switches.

James Ferrell and Eric Machleder incubated *Xenopus* oocytes, immature frog eggs, in different concentrations of progesterone, the hormone that starts their maturation. Each egg is about 1.2 millimetres across, large enough to assay alone by [immunoblot](https://stemhop.com/methods/western-blot/), a gel-based protein detection method in which active MAP kinase runs slightly slower. Pooling eggs at each dose produced a smooth S-shaped average.

**Predict.** Which experiment distinguishes two possibilities: every egg is partly active, or some eggs are fully active while the rest are off?

> **Check your prediction.** Assay one egg per lane. A pooled average cannot distinguish intermediate cells from a mixture of fully on and off cells.

The single-oocyte assay supported the second story. Ferrell and Machleder reported on 8 May 1998 that the response had a Hill coefficient of at least 35. The **Hill coefficient**, *n*, measures how abruptly output rises with dose. Hemoglobin binding oxygen, with *n* about 3, is the textbook comparison.

> **response = dosen / (Kn + dosen)**, where K is the dose that gives half the maximal response
>
> **dose at 90% of maximum ÷ dose at 10% of maximum = 811/n**

**Predict.** Use the second relation. How wide is the 10-to-90 percent dose span for *n* = 3 and *n* = 35?

> **Check your prediction.** The spans are 811/3, about 4.3-fold, and 811/35, about 1.13-fold. For *n* = 1 the span is 81-fold. Larger *n* means a narrower, more switch-like response.

Extracellular signal-regulated kinase (ERK), a MAP kinase, turns on only after MAPK/ERK kinase (MEK) adds two phosphates, one to threonine and one to tyrosine one residue apart. One tag leaves ERK mostly off. MEK adds the phosphates in separate encounters, releasing ERK between them. This multisite process makes the response **ultrasensitive**: steeper than proportional, an S-shaped curve rather than a true threshold, and still reversible. Stacking three relay tiers sharpens it further. On 17 September 1996, Chi-Ying Huang and Ferrell published a calculation, confirmed in oocyte extracts, with a Hill coefficient about 4 to 5.

Thirty-five needs more. **Positive feedback** means that output increases its own production. Ferrell and Machleder proposed that a loop around the cascade drives an egg that has started switching the remaining distance. Sufficiently strong positive feedback can add **bistability**, two stable output states under the same conditions, and memory. Xiong and Ferrell showed in 2003 that an oocyte committed to maturation stays matured after hormone removal.

Eggs differ in switching threshold, so at an intermediate dose a homogenate reports the fraction that has flipped as if every egg responded partway.

*Figure (AI-generated illustration).* The same population average can hide partly active cells or a mixture of fully on and off cells, so single-cell measurements are decisive. Image: two rows of four circles (eggs) show two stories that give the same pooled reading. Story one, every egg partly on: four uniformly pale-gold eggs feed a tube labeled Pooled and then a pale-gold reading box. Story two, all or none, marked with a check: two dark-gold eggs labeled Fully on and two empty eggs labeled Off feed a tube and an identical pale-gold reading box. A callout reading Same pooled reading connects the two boxes. A bottom strip, One egg per lane, shows a single egg loaded into one of four empty gel lanes.

## Adaptation: returning to rest while the signal is still on

**Adaptation** is a return toward rest while the stimulus remains. It lets *E. coli* respond to a climb in attractant rather than stay saturated by the level already reached. The second film showed receptor methylation acting as a running total of the error, which engineers call **integral feedback**, a negative feedback that keeps correcting until the error is zero. Here is the molecular wiring.

The receptor controls CheA, which transfers a phosphate to CheY; phosphorylated CheY drives tumbling. CheR adds methyl groups to the receptor and CheB removes them. The idealized model works because CheB acts mainly on active receptors and CheR works near its top speed. Like turning a shower knob until the water feels right, correction speed affects arrival time, while “right” sets the stopping point.

In most networks, changing an enzyme’s amount shifts the steady state the network settles into. Barkai and Leibler derived in 1997 that this loop should instead return each cell to its own pre-stimulus output across wide ranges of enzyme levels. On 14 January 1999, Alon, Surette, Barkai and Leibler tested *E. coli* carrying different amounts of one circuit protein at a time. Adaptation time and steady-state behaviour varied strongly with protein concentration. The precision of adaptation did not. **Robustness** means that a specified output property holds when component amounts or rates vary; it never means that every property is protected.

With an overloaded correcting enzyme, a cell can still return within experimental error to its own pre-attractant rate while differing from normal neighbours in resting rate and return time.

**Predict.** Each receptor has a fixed number of methylation sites. What should happen at very high attractant levels?

> **Check your prediction.** Free methylation sites run out before the error reaches zero, so real cells adapt measurably short of their starting rate. Exact adaptation is a limit of the model, not a guarantee.

## Put the circuit words to work

| Circuit word | Question it answers |
| --- | --- |
| **Gain** | How many output molecules does one input produce? |
| **Negative feedback** | Does output act back to reduce its own production? |
| **Positive feedback** | Does output act back to increase its own production? |
| **Feed-forward** | Does one input reach an output by direct and indirect routes at once? |
| **Ultrasensitive response** | Does output climb from near zero to near maximum over a narrow input range? |
| **Adaptation** | Does output return toward rest while stimulus remains? |
| **Robustness** | Which specified output holds as parts or rates vary? |
| **Insulation** | How is one pathway’s signal kept from another’s targets? |

**Cross-talk** is signal passing between pathways. A cell may use that connection, or it may be an unwanted leak. Chris’s lectures borrow a picture from process control: water, malt and hops flow into a fermenter, a threshold indicator trips, and product goes to a storage tank. Two of his questions apply to every case here: “Insulation versus connectivity?” and “Does a signaling pathway have to be the most logical or efficient of all possible scenarios?” The chemotaxis circuit answers the second. It is not the simplest wiring that responds; its architecture and kinetics protect one output, how precisely each cell returns to its own starting rate, and not everything.

## Going deeper

### Different inputs, different activity windows

One summary figure attributes feedback-shaped activity windows of about 20 minutes after TNF (tumor necrosis factor), 30 after EGF (epidermal growth factor), 90 after NGF (nerve growth factor) and 120 after LPS, a molecule from bacterial outer membranes, to the loops handling each signal. These examples do not make feedback the only possible timer.

### What the simple gain product leaves out

The product rule counts independently multiplying steps before saturation. Real pathways may share substrates, saturate enzymes, deactivate components while transmission continues, and feed output back into earlier steps. Rate multiplied by active lifetime helps set a step’s gain, but amplitude and duration are not locked together: downstream shutoff or feedback can shorten a high-gain output.

### Exact adaptation is a model limit

Experiments show return within measurement limits over a range of protein amounts, not unlimited perfection.

---

From STEMhop (https://stemhop.com), free lessons by Christopher Deppmann. Lesson text © the authors. Practice questions: https://stemhop.com/lessons/signaling-circuits/
