Bionics News Digest

A robot with a bacterial brain, a spinach sniffer and other bionic tricks of October

Mikhail Petrov, "The Attic"
(for links to publications and explanations of terms, see the original article – VM).

Implanted electronics, chimera robots and biological computers are increasingly blurring the vague boundary between living beings and inanimate matter. The Attic will select the most interesting and important news from the world of bionics so that you don't get lost in this new, strange world.

How to grow a minesweeper

Researchers from the Massachusetts Institute of Technology (MIT) have taught spinach (Spinacia oleracea) to look for explosives. To do this, they pumped into its leaves a solution of carbon nanotubes with bombolitin protein molecules wound on them – this mixture is able to recognize nitroaromatic compounds, the basis of many explosives, due to the fact that in the presence of nitroaromatic protein changes its spatial structure. According to the domino principle, this protein transformation leads to a slight rearrangement of the electronic levels of the nanotubes associated with it, which changes their radiation spectrum.

In experiments, scientists added trinitrophenol to the soil – the plant absorbed it along with water through the roots, and then the explosives gradually rose along the stem to the leaves. There it interacted with nanotubes, and to read the signal (those characteristic changes in the spectrum), scientists illuminated spinach with a laser, which excited the fluorescence of nanotubes in near-infrared light (a small infrared camera controlled by a Raspberry Pi microelectronic board was used to register the fluorescence spectrum).

Spinach-sniffer found trinitrophenol in an average of eight minutes, and now researchers dream of using it or other similar nanobionic plants for constant monitoring of soil condition: their branched root system can effectively "collect data" over a large area, and sensitive sensors built into the leaves will not miss any contamination. However, it is still unclear how to make plants not only look for something other than niatroaromatic compounds, but also be able to distinguish different toxicants.

The harm of smoking was assessed using an artificial lung

All modern methods of assessing the harm of smoking have their drawbacks: testing of cell models grown on human cells is incorrect, because these models do not know how to shrink and straighten out, as our lungs do; tests on mice, rabbits and other laboratory animals are questionable, because all these animals breathe through their noses, and not finally, human studies will never allow us to compare the condition of the same person in the states of "a heavy smoker" and "has not smoked a single cigarette in a lifetime" (scientists have to compare different groups of people and catch the changes associated with smoking against the noisy background of individual characteristics of each person).

Therefore, scientists from Harvard (Wyss Institute of Biologically Inspired Engineering) forced their new, most realistic lung model to smoke, while they themselves recorded in parallel what genetic and physiological changes this leads to.

Their artificial lung looks like a plastic chip with a hollow channel seeded with bronchial epithelial cells, for whose life conditions are almost identical to physiological ones – the same pressure, temperature, humidity and periodic mechanical stresses as in the lung. On the one hand, this channel is in contact with air, and on the other – with a porous membrane that separates it from the flow of epithelial cells that constantly renew the tissue of the model.

To simulate smoking, the Americans also made a separate device, smoldering cigarettes and releasing smoke, not anyhow, but as realistic as possible – puffs. In the first experiment, scientists looked at which genes were activated or, conversely, deactivated in epithelial cells after smoking (that is, after the smoke of an artificial smoker was passed through the channel instead of air), and found changes in 335 genes involved in a total of 23 different biological processes (for example, inflammatory processes or cell adhesion).

At the same time, the expression of the actioxidant gene HMOX1 increased the most – it increased by more than 15 times, and therefore, in the next experiment, the scientists decided to see if such an effect would be repeated for e–cigarette smoke, but they did not see anything like this - the expression of HMOX1 did not significantly change the expected harm (at least, according to these data). they didn't see it.

Finally, in two more experiments, we compared how smoking affects gene expression in healthy cells and cells suffering from chronic obstructive pulmonary disease COPD (147 specifically activated genes were found in artificial lungs of COPD patients), and looked at how cigarette smoke changes the nature of the movement of cilia covering the bronchial epithelium and cleansing it of impurities (movement the cilia became inconsistent: if before exposure to smoke, the frequency distribution of their oscillations fell on the Gaussian distribution, then after this dependence disappeared).

A robot with bacterial brains

We used to think that only the nervous system is responsible for information processing and decision–making in our body, but this is not so - the bacteria inhabiting the insides of a person also synthesize neurotransmitters and are quite capable of influencing our mood or way of thinking (in science, even now there is a term gut–brain axis - "brain-gut axis"). Therefore, scientists from Virginia Polytechnic University (USA) have assembled a robot whose movements are controlled not only by programmed electronics, but also by bacteria, which, according to the authors, should bring their robot closer to human thinking, collected at least from brain signals and intestinal signals.

The robot itself looks like a small cart with two wheels, a video camera and a motor. She can drive forward and turn right and left, and her movements are controlled by an Arduino board and a culture of genetically modified E. coli, which lives in a separate, static tank and exchanges information with the robot via Bluetooth.

Bacterial brains work as follows: when two pathogen chemicals (lactose or arabinose) enter the culture, they trigger a cascade of biochemical reactions leading to the synthesis of one or two fluorescent proteins glowing green (green fluorescent protein) or red (mCherry protein). A separate camera captures the intensity of this glow and transmits the data to the trolley, where they are already processed by Arduino and affect what decision (to go forward, stop, turn right/left) the robot's electronic brains will take at the next moment.

Scientists tested their design on a miniature arena, at different points of which green and red cylinders of the same size were placed, providing feedback from electronics to bacteria: a camera on a trolley took a picture, built-in computer vision algorithms recognized how much red or green color came into the robot's field of view, and transmitted this information to the culture of microbes, where she set the input levels of concentrations of the very substances-pathogens that trigger bacterial calculations.

Test data from one of the previous versions of the biorobot. Orange rectangles indicate the sources of arabinose, and blue triangles indicate the sources of lactose, between which the robot is looking for the optimal path. Source: Scientific Reports

In the work, the scientists set their biorobot tasks to search for red and green cylinders and looked at which strategies for pairing signals of electronics and bacteria would cope with them best. It turned out that each of them has its advantages: for example, the robot coped with the search for a single green cylinder most quickly on the principle of master-slave (the leader was a culture of bacteria, the slave was electronics: here the feedback between the two systems was turned off), and the robot controlled by an algorithm with included feedback, which scientists compared with hormones synthesizing under the control of the nervous system and affecting the microbiota.

In addition, some strategies allowed the biorobot to even show the simplest signs of adaptive behavior – for example, he was able to change his goals on the go without making changes to the program (for example, he first looked for green cylinders, and then red) or, conversely, ignored a certain type of goals in favor of his "favorite" prey (in search of a remote red the cylinder could drive past the closely located greens).

Neuromorphic computers based on polymers

A memristor is another, fourth passive element of microelectronics, along with a resistor, a capacitor and an inductor. Its main feature – the dependence of the conductivity of the memristor on the charge flowing through it, that is, on the history – makes it fundamentally similar to synapses – contacts between neurons (or between neurons and cells) that transmit signals the better, the more often they have to do it. That is why, based on memristors, they hope to assemble neuromorphic processors capable of repeating biological calculations in hardware, and not just imitating them with the help of clever program code, as modern neural networks do.

Memristors can be made on the basis of a variety of materials, including the electrically conductive polymer polyaniline (PANI), in which a small external electrical voltage causes an electrochemical reaction that changes the structure of the polymer, its color and, most importantly in the case of a memristor, conductivity. Previously, a single–layer perceptron was already assembled on the basis of memristors - the simplest neural network capable of not only solving certain computational tasks, but also learning in the course of its work, and now Russian scientists, together with Italian colleagues, for the first time made a two-layer perceptron using organic memristors.

In a two–layer perceptron, as the name suggests, there are several layers: the first of them, the sensory, receives external signals from other fragments of networks (and there are no neurons in it - elements capable of learning); the second, the associative layer, processes these signals (there are neurons capable of rebuilding depending on the passing information); finally, the third, reacting layer, issuing signals back to the external circuit (there are also neurons here, and that is why the perceptron is called a two–layer - the sensory layer without neurons does not count)

In the development of Russian and Italian scientists, the two-layer perceptron circuit was assembled using 12 memristors, MOSFETs and other microelectronic components. As a result, their perceptron can be trained to solve the exclusive OR problem – that is, it can recognize those cases when opposite signals are applied to its input (this is useful for classifying objects and single-layer perceptrons cannot cope with similar tasks), and can also cope with the simplest analog tasks (classify signals other than logical 0 and 1).

Used articles Benam, K. H. et al.
Matched-Comparative Modeling of Normal and Diseased Human Airway Responses Using a Microengineered Breathing Lung Chip. Cell Syst. 3, 1 (2016).
Emelyanov, A. V et al. First steps towards the realization of a double layer perceptron based on organic memristive devices. J. Appl. Phys. Proc. 6, 111 301 (2016).
Heyde, K. C., Gallagher, P. W. & Ruder, W. C. Bioinspired decision architectures containing host and microbiome processing units. Bioinspir. Biomim. 11, 56 017 (2016).
Wong, M. H. et al. Nitroaromatic detection and infrared communication from wild-type plants using plant nanobionics. Nat. Mater. (2016).

Portal "Eternal youth" http://vechnayamolodost.ru  16.11.2016