Almost one hundred and seventeen years after it was first proposed, we now know that single-celled organisms can make decisions.
Can a single-celled organism make decisions and learn from its environment?
Can it have a mind of its own?
More than a century ago, a biologist, Herbert Spencer Jennings, observed so. However, his observations were later ignored as they were supposedly not replicated.
Then three system biologists, two from Harvard and one from India, decided to do Skunkworks testing of the discarded paper and show that a single-celled organism does seem to have an inner capacity that is quite remarkable.
After a century, Jennings stands validated and how!
The Single-Celled Organism
Think about a single-celled organism and what is the image that you get?
An amoeba or a paramecium — engulfing their food in a water-filled environment, primitively sensitive to light and moving around with the very basic instincts of life.
That is how most of us ‘see’ them through our textbooks and also sometimes through the microscopes.
Perhaps, they spread disease and contain biochemical cures hidden in their chemistry. If we go a little deeper, we may find their role in ecology — in the natural cycles of the planet.
In the large drama of planetary evolution, we tend to see them as background actors. Naturally, since they have the most primitive sensory mechanism and possess arguably the simplest way of processing the information they get from the environment around them.
But, do they have something more complex than we assumed?
One hundred and seventeen years ago in the ‘American Journal of Physiology’, a paper was published by one of the pioneers of American biological sciences, Herbert Spencer Jennings. The paper studied how single cellular organisms reacted to the stimuli provided.
Jennings, then an established name in the study of a particular branch of unicellular organisms called ‘protozoans’, had shown in his paper how the organism exhibited a complex series of responses.
Of all the organisms available to him among the ‘Protists’ as these single-celled animals are called, he chose a sessile ciliated organism.
To him "the behavior of an animal which is fixed in a definite position will necessarily be of a different character from that shown by such an organism as Paramecium ... showing indeed a much higher development."
Further, such an organism has the advantage of keeping "the same individual continuously under observation, or return to it at longer or shorter intervals" which in turn makes it "possible to observe changes in behavior and to determine whether the reaction to a given stimulus is modified by previous subjection to the same or different stimuli" and through such criteria he found Stentor roeselii “in many respects the most favorable as well as the most interesting of the organisms studied”.
Stentor oeselii is colourless and has a trumpet shaped body. At the one end of its body is a disc like expansion. This is called the peristome.
While there are longitudinal cilia rows covering the body, the ending disc has a circular arrangement of cilia called peristomal cilia or membranellae.
The animal is sessile usually when feeding, attaching itself through minute pseudopodia to a submerged body, mostly on a plant surface or decaying organic matter.
The peristomal cilia create a vortex in the water leading to the mouth along with particles near the axis of the vortex.
Stentor Roeselii And Stimuli
When the animal was given stimulus, it was observed by Jennings, to be producing a variety of responses, actually five types of responses.
These differing responses were for the same stimulus and observed to be appearing in a sequence.
The stimulus was: releasing a large quantity of fine particles, such as India ink or carmine, by means of a capillary pipette into the mouth of an organism that has already attached itself to an object.
The initial response is being passive.
Then it bends to one side, toward the aboral side, that is away from where the oral cavity was. This is effective if the number of particles released are not high.
However, if the particles are continued to be released, after one or two times of trying this, the animal switches to the next behaviour of defence.
Observed Jennings, "the ciliary movement is suddenly reversed in direction, so that the particles against the disk and in the pouch are thrown off. The water current is driven away from the disk instead of toward it."
After trying this, it then again allows the water to flow in the normal way. If the irritation of the stimulus continues, this response is also repeated once or twice and then the animal moves to its next defence behaviour — complete contraction.
The organism contracts into its tube. While this helps Stentor to escape the carmine particles completely it also means complete cessation of all activities — importantly, obtaining food.
So, after half a minute the organism now starts expanding. As the animal reaches two third of its original height, the peristomal ciliary action begins as before, directing the water and particles towards its mouth.
If the irritation is still persistent then how will the organism behave? Will it begin the entire defence cycle again?
Now, Jennings asked the crucial question: "Or shall we find that it has become changed by the experiences it has passed through, so that it will now contract again into its tube as soon as stimulated?"
The animal does not repeat the response scheme from the beginning. But it contracts into the tube again. And extends again. If the stimulus is still there then each contraction stays a little longer in the tube than it did at first.
At last it only contracts "repeatedly and violently", observes Jennings. This in turn has the effect of freeing the attached 'feet' from the surface to which it is attached. Then the organism leaves the tube and swims away to another location away from the stimulus and attaches itself there.
For Jennings, the observations had very important implications for our understanding of how living organisms behave.
Pointing out that "the anatomical structure of the organism and the different physical or chemical action of the stimulating agents are not sufficient to account for the reactions”, he decided that the change in the internal state of the organism should account for its variety in response.
The reaction to any given stimulus is modified by the past experience of the animal, and the modifications are regulatory, not haphazard, in character. The phenomena are thus similar to those shown in the 'learning' of higher organisms, save that the modifications depend upon less complex relations and last a shorter time.
Jennings ‘Refuted’, Or Was He?
But in 1967, more than half a century later two researchers, James Reynierse and Gary Walsh would contest these claims. They tried to replicate the ‘behavior modification' in Stentor "by using a classical conditioning procedure".
They concluded that "regardless of experimental conditions, whenever carmine particles were administered to Stentor, it quickly became free-swimming" and hence declared that "the phenomenon of learning was not demonstrated in this experiment".
Soon, Jennings’ findings were forgotten.
Jennings Proved Right
And then, in 2019, three systems biologists, two from Harvard University, Jeremy Gunawardena and Joseph Dexter, and Sudhakaran Prabakaran from IISc, India, now with Cambridge University and IIESR Pune, decided to revisit Jennings’ paper.
With no conventional support coming for revisiting a forgotten paper almost a century old, the trio had to work nights.
One of the reasons for revisiting the findings of Jennings was because Gunawardena found the 1967 paper "one of the shoddiest studies" he had ever seen.
The organism used was not the one Jennings used. It was a different species. This was because of the then dominant scientific paradigm reigning generally in psychology and studies of animal behaviour — behaviourism where one organism was as good as the other.
But as Gunawardena says "paradigms change, as happened when cognitive science replaced behaviourism, but it can take a long time to rescue what was previously marginalised”. And in this case more than 110 years.
Gunawardena and his team decided to use the exact organism Jennings used. But when the experiments were attempted there was initially disappointment.
The organism did not respond at all for the carmine stimulus. Then when in desperation the team started to use a few other substances they found that microscopic polystyrene beads (red-fluorescent latex beads) in Sodium azide (NaN3) solution, produced all the responses Jennings had observed.
Clearly carmine composition had changed in a century. They also did video-recording of the experiments.
They not only got the different avoidance mechanisms which Jennings had observed but also more heterogeneity in responses than he observed.
But is there a response hierarchy as concluded by Jennings? The team took a statistical approach to the question.
They conducted 60 separate experiments with one, two, or three organisms tested with beads stimulation of one to seven pulses and detailed tables were made and analysed.
The paper now clearly proves that there is a hierarchy.
For example, ciliary action (which they labelled as 'A’ for statistical analysis) and bending away (labelled ‘B’) always happened together. So, these two responses bundled together as 'A or B' always happened before detachment from the attached surface (labelled D). Similarly, contraction (C) always happened before D.
So, there is indeed a hierarchical series of behaviour in a hierarchical manner.
Should we also now think of the organism as capable of ‘decision making’ — a single celled organism?
There is for example the question of ‘Clever Hans’ effect. The effect was originally named after a horse that was famous for its supposed ability to solve complex mathematical problems which on investigation turned out to be the horse giving the right answers by picking up cues people were giving out unaware.
In this context, the Clever Hans effect refers to if "the experimenters, had subliminally learned how to elicit the complex behaviors they were seeking".
The paper rejects the presence of such a bias effect by pointing out two aspects, one: two individual organisms with the same stimuli eliciting different responses and the very existence of a behaviour hierarchy.
While the paper has definitely rediscovered and validated the discovery of Jennings it also increases our understanding of the dynamics of life at the cellular level.
The paper states:
The results presented here confirm that Jennings was right. ...We find substantial heterogeneity in behavior, which Jennings did not address, but by following a quantitative approach, in contrast to his descriptive methods, we provide compelling evidence for Jennings’ behavior hierarchy. Remarkably, the choice between contraction and detachment is consistent with a fair coin toss, raising the intriguing question as to how S. roeseli implements this so accurately at a molecular level.
No wonder, Dr Chaitanya Giri, planetary and astromaterials scientist, who has worked for over a decade at the Earth-Life Science Institute at Tokyo Institute of Technology (not involved in the study) considers this paper "by far one of the most important discoveries transcending between two grand human quests; that of knowing the origin of life and the origin of awareness".
Then there are new windows being opened — or should one say a rabbit hole descent into a wonderland.
The paradigm shifts the paper hints at also extend beyond study of the behaviour of single-celled organism.
The paper suggests the membrane of Stentor as well as its cytoskeletal cortex as 'the most likely candidates for mechanistically implementing the behaviors observed here.' They have some interesting characteristics which they share with neurons that would evolve long after: 'The ciliate membrane is excitable. It harbors voltage-dependent and mechanosensitive ion channels that generate action potentials' making them 'analogous to those in neurons'.
So the organism while not having neurons, does have a neuronic ability.
What is even more interesting is that the cortex can pass on to the daughter cells through 'a non-genetic and Lamarckian manner' minute alterations which happen to the nano-geometry of the cilia - a kind of 'cortical inheritance.'
The word Lamarckian has become a kind of bad word in biological sciences, particularly after the political abuse by Stalin's USSR and the pseudo-science of Lysenko. However with epigenetics and other discoveries the word is now no more carries the stigma it carried.
'Indeed, Lamarck is no longer a prohibited name in biological discourse' says Dr. Jeremy. Even 'the central dogma' of molecular biology (strictly DNA->RNA->protein arrow) has become 'so modified that it is barely recognisable', he points out. But then could have such mechanisms played a role in evolution also? That still remains very controversial, cautions Dr.Jeremy while he himself thinks that the evidence is slender for a strong evolutionary impact of such mechanisms.
The story does not end here.
Dr. Sudhakaran Prabakaran points out to a subsequent paper published in i. Here along with Dr. Sudhakaran two more scientists from University of Cambridge (Mi Kieu Trinh & Matthew Wayland) made a 'behavioural analysis' of Stentor roeseli using 'machine learning approaches.' The first model was based on the decision tree algorithm. The second model used was random forest with multiple trees being generated instead of the single tree approach which conventional decision trees use. Further the investigation was done by artificial neural networks feed-forward neural networks made of structured computational layers. These layers consist of computational 'neurons' called 'perceptrons' which in turn mimick 'actual biological input and activation architecture of real neurons'. However, the results show that such an artificial neural network with multiple ‘computational’ neurons is inefficient at modelling the single-celled ciliate’s avoidance reactions.
The ANN models used have five perceptron as input layer, each representing an input which are being at rest; bending; cilia movement; contraction time and contraction number and then there were varying 'hidden layers' each with varying perceptron in three ANN models used. One with one hidden layer with six perceptrons and another with 2,10 and 6 perceptrons in each of the three hidden layers and yet another with 2 hidden layers with 9 and 18 nodes each. It was found that it was the third model that came close to some success (59 percent accuracy) while the simple first model was not at all of use. The researchers observe the following about the results:
These results suggest a high level of complexity in the behaviours of Stentor roeseli in response to external stimulation. It cannot be fully explained by habituation, sensitization or operant behaviour. Our machine-learning based models, though impressive in modelling activities of neural systems like the brain, have been largely unsuccessful when applied here.
‘Biology has many examples of "computation" besides neuronal networks’ Dr. Gunawardena points out, like ‘ecosystems, immune systems, gene regulatory networks, post-translational modification systems’ etc. They all can ‘implement more or less complex forms of information processing at different biological scales.’ Dr. Giri also points out that ’a thrust on understanding natural intelligence’. According to Dr.Giri the reason present-day algorithmic approach will not work for unicellular organisms: ‘I am afraid, as most algorithms are known to work under unambiguous factors - whereas Stentor roeseli has shown tremendous ambiguity in decision-making.’ The paper proposes that more interconnected Boltzmann machines.
For Dr.Sudhakaran the critical question of the study is 'can invertebrates learn?' For this we need to have a taxonomy of learning, he says in the paper. In this regard he cites Dr.James V. McConnell (1925-1990) who known for his experiments with invertebrate learning. In 1966 Dr. McConnell had lamented 'that there is no ‘systematics’ in behavior and that comparative psychology is still awaiting its Linnaeus.' While we are talking about 'learned responses' and 'unlearned responses' the question what learning itself is seldom explored. And such a true definition of 'learning' in a single-cell organism has to explain 'both 'innate response' and Enactivism'.Innate response sees the organism as having already endowed by evolution 'a highly complex pattern of ready-made responses' and so here ‘learning’ is what the environment imposes and elicits from this internal system with which organism comes. Enactivism endows the organism with the ability to create 'its own reality through dynamic interaction with their environment, assimilating information about the outside world into their own ongoing dynamics, not in a reflexive way, but through active inference, such that the main patterns of activity remain driven by the system itself'. Dr. Sudhakaran adds enigmatically, 'you can call it consciousness or free will.'
- Dexter et al., 'A Complex Hierarchy of Avoidance Behaviors in a Single-Cell Eukaryote', Current Biology (2019), https://doi.org/10.1016/j.cub.2019.10.059
- Mi Kieu Trinh, Matthew T. Wayland & Sudhakaran Prabakaran, 'Behavioural analysis of single-cell aneural ciliate, Stentor roeseli, using machine learning approaches', Journal of The Royal Society Interface (2019), Vol.16, Iss. 161 https://doi.org/10.1098/rsif.2019.0410
- · Herbert Spencer Jennings, 'Behavior of the Lower Organisms', Columbia University Press, 1915
- · Kevin Jiang, 'Unexpected Depths: New study hints at complex decision making in a single-celled organism', 'News and Research', Harvard Medical School, Dec-5 2019.
Swarajya thanks system biologists Dr. Jeremy Gunawardena & Dr. Sudhakaran Prabakaran (both involved in the study) and astrobiologist Dr. Chaitanya Giri (not involved in the study) for their time in interacting and responding to our queries. All the views as they came out in the interactions with respect to these two papers from both Dr. Jeremy Gunawardena and Dr. Chaitanya Giri require separate reproduction in their entirety.