A principle of organization which facilitates broad Lamarckian-like adaptations by improvisation

Examination of the rapid world-wide increase in antibiotic tolerance of pathogenic bacteria (commonly thought to be associated with acquisition of resistance conferring genes [87, 88, 92, 93]) reveals all the hallmarks of the proposed principle. Bacterial genomes are subdivided into a “core” genome arranged in chromosomes, and an “accessory” genome consisting of mobile genetic elements (MGEs) such as bacteriophages, pathogenicity islands, chromosomal cassettes, plasmids and transposons [94, 95]. These MGEs can be viewed as “weakly-constrained features” capable of contributing to de novo adaptation, which can be inherited by vertical and horizontal gene transfer (HGT) mechanisms [59, 92, 96–98]. Each MGE can also have hundreds of non-identical copies and can account for 25 % of the total DNA content [99]. Unlike the core genome which contains all the constitutively vital genes (metabolism, DNA synthesis etc.), the accessory genome is enriched with genes promoting adaptations to various ecological niches (resistance, virulence factors etc.) [100, 101]. Genetic analysis at a population level shows that MGEs are more variable compared to most regions of the genome [102, 103]. This variability is thought to reflect mutagenic potential [104], high recombination rates [105] and other forms of genetic instability and intercellular exchange of genetic elements [106–108]. Moreover, changes in MGEs or their complete loss, is typically less detrimental to the bacteria compared with chromosomal loss. Accordingly, several studies suggest that HGT of MGEs is a major factor in regulatory evolution [86] and a key contributor to acquisition of bacterial tolerance to toxic stress [92, 97, 109, 110].

Notably, some of these transfer mechanisms are increased or enabled by stress [111], and various classes of MGEs also exhibit stress-dependent increase in the recombination and mutation rates [104, 105].

The acquisition of bacterial stress tolerance by random changes in MGEs and their exchanges between individuals, is also compatible with stabilization of tolerance by positive (auto-regulatory) feedback mechanisms. For example, nutrients and signaling molecules that are induced (directly or indirectly) by beneficial changes in bacterial MGEs, may promote their own production in the altered bacteria as well as in their non-altered neighbors.

In this particular example, adaptive improvisation relies on two distinct carriers of genetic information: A relatively stable, slowly evolving genome which encodes many essential functions, and a flexible genome which is more sensitive to stress, varies on faster timescales, exchanges more readily between bacteria and has reduced potential to harm the bacteria (compared with equivalent changes in the core genome). Under new stressful conditions, bacterial MGEs can undergo many rapid changes, which may alleviate the stress and reduce the drive for further changes.

Adaptation by improvisation does not necessarily rely on genetic changes. One example based on diverse types of potential tolerance-conferring factors is provided by production and exchange of exosomes. Many cells release small (30-250 nm sized) extracellular vesicles which can be loaded with diverse set of biomolecules, including DNA, RNA (both coding and non-coding), proteins and other cytosolic components [112, 113]. These vesicles can deliver factors which support growth [114] and enable inter-cellular communication [115, 116], which could allow coordination of cellular activities. Exosome release can be enhanced by stress [117], increase stress tolerance [118–120], assist in promoting or evading immune responses [121, 122], facilitate tumor survival and progression [123–125] and may also protect cells from pathogens [114].

Production and exchange of exosomes between cells increases the overall amount and diversity of changes. Moreover, when a subset of cells produces exosomes with factors that contribute to stress tolerance, the exchange of these exosomes increases the tolerance of recipient cells. If the underlying cargo is also self-enforcing, the increase in tolerance could be substantially enhanced and stabilized. Participation of a large population in exosome production and exchange further allows each cell to acquire tolerance with a lower level of exploration on its part. This benefit of cross tolerance reduces the potential for negative outcomes. As in the case of MGEs, the potential for horizontal transfer of DNA via intercellular exchange of exosomes provides a straight-forward mechanism for transgenerational transfer of the acquired stress-tolerance [126]. The discovery of functional contributions of exosomes in marine ecosystems [114] suggests that they may account for part of the rapid adaptation of corals to climatic change [127].

A particularly notable (but not exclusive) example in multicellular organisms is based on the symbiotic microbiome. The latter typically includes a very large number of micro-organisms, capable of responding more readily and rapidly to external or internal perturbations (compared with host cells). The contribution of symbionts to the development and homeostasis of their hosts has long been recognized [52, 128–130], and more recently referred to as the holobiont theory [131–133]. While this theory also recognized the potential contribution of heritable symbionts to rapid evolution of their host, it did not address the possibility (or requirement) for frequent emergence of de novo adaptations in each individual. Accordingly, it does not specify an organizing principle by which random changes in symbionts can account for de novo emergence of individual–specific adaptations during the lifetime of evey individual. We propose that the broad host-microbiome interactions enables stress reduction in the host by random changes in the microbiome. Under stressful conditions which are not efficiently alleviated by pre-evolved responses, the microbiome rapidly undergoes a series of diverse changes (e.g. changes in species composition, gene sequence, epigenetic state, etc.). These changes could be induced by direct exposure of the microbiome to new external environments or indirectly by an input from the stressed host [134]. In a very short period of time (relative to the generation time of the organism), a very large number of micro-organisms can change and produce (or eliminate) many products which could affect the host. If the alterations in the microbiome happen to alleviate the stress, they also reduce the drive for further changes and the new microbiome state is more likely to persist/stabilize (Fig. 3).

In this way, the microbiome can act as an adaptive buffer, providing the organism with at least partial solution that reduces negative outcomes for the host and diminishes the drive for varying more stable host-intrinsic traits. It is important to note, however, that induced changes in the microbiome are not guaranteed to reduce the stress. In many specific scenarios, the stress will either remain at similar levels or even increase. However, in these cases, the microbiome will likely continue to change, thereby exploring new avenues for mounting a beneficial response. This should create a statistical bias towards better outcomes compared to those achieved without the feedback, resulting in net probabilistic gain of adaptation within the lifetime of a single individual.

As in the cases of MGEs and exosomes in single cells, beneficial changes in the microbiome of animals can be amplified and stabilized by feedback mechanisms. Random changes in bacteria which happen to reduce the stress and also promote expansion of the altered bacteria (on the expense of other micro-organisms), could lead to further expansion of the beneficial bacteria and reciprocal displacement of less beneficial micro-organisms. The beneficial outcome can also spread in a population of hosts via the transfer of these bacteria.

Regardless of the eventual outcome on the stress, changes in the microbiome affect host-intrinsic processes [128, 131–134]. As such, the changes in the microbiome can promote physiological and epigenetic modifications in the host, which in turn feedback on the microbiome until the stress falls below a threshold for a change (or, alternatively, until the organism dies). Hence, a bacterial-mediated (adaptive) process of stress reduction is expected to induce a gradual change in the host, particularly in its more weakly constrained internal processes and components. We have recently hypothesized that the microbiome can assist the organism in being both robust and plastic by balancing stability and flexibility based on contextual demand [134]. The current proposal extends this idea by suggesting a mechanism in which the robustness of important traits is achieved by the plasticity of less stable ingredients, such as the symbiotic micro-organisms. This highlights the dichotomous contribution of plasticity to both generation and suppression of variation.

The potential heritability of adaptability-conferring changes in the microbiome, provides rich infrastructure for rapid emergence of new adaptive responses. This buildup of adaptive responses is greatly assisted by the accumulation of beneficial changes in every generation of the organism, thus improving the overall response within timescales that are much faster compared to genetic adaptation of the host [132, 134]. Potential examples include the rapid spread of the defensive endosymbiont Spiroplasma in Drosophila hydei exposed to high parasitoid wasp pressure [135], the regulation of thermo-tolerance in the Aphid A. pisum by mutations in its obligatory endosymbiont, B. aphidicola [136], the regulation of plant specialization in this Aphid by its facultative proteobacterium [137] and the rapid evolution of competitive, yet suboptimal strains of symbiotic mesorhizobia on the legume B. pelecinus [138].

While the microbiome provides a relatively simple and powerful realization for acquiring new adaptations in metazoans, the proposed principle of stress-regulated improvisation in multicellular organisms is not limited to changes in the microbiome. It is broadly applicable to other host-intrinsic factors and processes. Similarly, the MGE- and exosome-based implementations in single cells should not be considered unique and likely occur in parallel to a variety of other mechanisms (e.g. prion-based acquisition of beneficial phenotypes under stress [59]).

Below we provide a set of predictions derived from various aspects of the hypothesis. It should be realized, however, that some predictions are potentially more informative, especially those which are designed to test if highly similar conditions of novel stress (e.g. in replicated experiments) can result in substantially different adaptive outcomes. Additionally, the dependency of adaptive improvisation on random processes, limits the ability to predict the outcome of every experiment. Accordingly, we use the terms’ expectations’ and ‘tendencies’ in a statistical sense to indicate that a particular prediction should be confirmed by averaging over a sufficiently large set of conditions and experiments, but is not necessarily expected in every setup or every experiment.

All living organisms have a capacity to reduce stress by improvisation within the lifetime of an individual. This capacity co-exists and can be influenced by pre-evolved responses. The rate and extent of the improvisation part of the response is expected to increase with the strength of the stress, which in turn, depends on the state of the organism, its current response and the external conditions. Compromising stress-relieving activities of pre-evolved responses (e.g. inhibiting heat shock response under exposure to high temperatures) will elevate the stress and increase the extent of random changes.

For example, organisms with a more diverse gut microbiome are expected to cope better with novel stress compared with the same organism with a limited diversity. Conversely, experimental removal of the microbiome (or reduction in other major sources of flexible variation) is expected to reduce the capacity to withstand novel conditions of stress. These conditions can be established by experimentally perturbing the endogenous regulation in ways which create highly unexpected maladaptive conditions (for example, by ectopic expression of a toxic gene from the promoter of an essential gene, or alternatively, placing an essential gene under the regulation of a silenced promoter [82, 139]).

The improvisation part of the response is expected to differ substantially between independent biological replicates of the same condition (in contrast to the reproducibility expected from pre-evolved responses). The probability of each outcome depends on the relative fraction of scenarios which lead to this particular outcome.

Initial support of this prediction has been provided by gene recruitment in yeast, showing that emerging adaptations ban be associated with global transcriptional changes which exhibit very little overlap between independent experiments [83, 84, 140].

For example, induction of a secreted auto-regulatory factor which promotes its own production in both the original producer and its neighbors will stabilize and synchronize the production across the interacting individuals. This coordination will tend to reduce differences between responses of co-exposed individuals compared with individuals in non-interacting populations that are subjected to similar conditions of stress. Similarly, adaptive responses which involve interactions between opposing changes can promote coordinated segregation into two anti-correlated subpopulations. Since the exact same rationale applies to interacting changes within a single individual, adaptive improvisation is also expected to promote correlations and anti-correlations between newly-forming changes within each individual (including within individual cells of a multicellular organism).

This prediction is also consistent with gene recruitment studies in yeast which revealed increase in gene expression correlations and anti-correlations under stress [83].

Beneficial outcomes of improvisation in a single individual (organism or cell) can be shared with its interacting partners, thereby acting to reduce the average stress per individual. This network contribution to stress-reduction decreases the drive for improvisation per individual.

Combining a wide range of adaptive changes that emerge by improvisation with mechanisms of non-Mendelian inheritance [141], creates a multitude of opportunities to transfer beneficial responses across generations. If newly forming adaptive changes are at least partly heritable, the stress is expected to decrease over successive generations of exposure to an ongoing stressful condition. For example, some of the changes in the gut microbiome which happened to have a positive impact on the host will be inherited by the host offspring. If the stressful conditions persist, the adaptation of the host will tend to improve by additional changes in the inherited microbiome. This processes will accompanied by a progressively reduced likelihood of complete reversal of the microbiome state and will eventually prevent reversal of corresponding traits in the host. While this example is based on changes in the microbiome, progressive improvement by coupling beneficial improvisation with non-Mendelian inheritance is widely applicable to host-intrinsic factors and does not depend exclusively on microbiome changes.

Because of the exploratory nature of adaptive improvisation, alleviation of stress by exploration will typically require a longer period of time compared to responses which have been previously selected based on their ability to accommodate this stress (e.g. pre-evolved responses to recurrent stress such as heat shock). A novel scenario of stress is also expected to delay ongoing processes. For example, a novel stress during a particular stage of animal development will tend to prolong this stage.

As a basic principle of biological self-organization, we expect the occurrence of adaptive improvisation in every stage and level of organization of living systems. These include the development, function and behavior of unicellular and multicellular organisms at any scale and arrangement (e.g. free living micro-organisms, symbionts, animal and plant cells, tissues and organs, complex societies of unicellular and multi-cellular organisms and larger ecological systems). Adaptive improvisation at any level can induce changes in any other level. For example, stress that is initially triggered in a specific cell within a tissue, will not necessarily remain restricted to this cell; a sufficiently strong stress will drive intracellular changes which affect nearby cells, thus leading to a tissue-wide stress. Likewise, a strong enough tissue-wide stress would scale up to the organ level. Similar scale-ups will lead to systemic, organism-wide changes which could further affect the external environment and the neighboring organisms. At every scale, the extent of adaptive improvisation will correspond to the degree and persistence of stress in this level. These interactions will tend to coordinate co-occurring variations in different locations and scales, leading to simultaneous stress reductions at multiple levels.

These predictions stem from the expected repressive effect of constraints on the ability to improvise. To maintain functional organization of a multi-cellular organism, the spectrum of deviations per cell needs to be narrower than in unicellular organisms. Without a sufficiently strong restriction (constraints) on the spectrum of deviations, adaptive improvisation in individual cells is more disruptive to coordination between cells, tissues and organs. This is much less of a problem in unicellular organisms, because the survival of each individual is less dependent on its degree of coordination with other cells. Unicellular organism are therefore more ‘free’ to adapt by improvisation compared to cells within animals.

Much of the constraints which restrict cellular changes that are harmful to a multicellular organism, are embedded in the pre-selected gene regulation of this organism. These constraints are compromised under conditions of dysregulation such as in tumorigenesis. Removal of constraints in these cells increases their capacity to adapt by improvisation and reduces their coordination with other cells in the tissue. This makes these cells less cooperative with the functional integrity of the whole organism. In that respect, tumor cells behave like unicellular organisms [58], which undergo rapid and relatively uncoordinated adaptations to new environments within the whole organism, eventually leading to its destruction. Since adaptive improvisation can be achieved by a wide variety of mechanisms, it provides tumor cells with an unusually broad adaptive potential which is not exclusively dependent on a single mechanism. This can account for the ability of tumors to resist a vast array of cancer treatments. We therefore expect a much larger efficacy of treatments which combine anti-tumor drugs with factors that limit the capacity of cells to adapt by improvisation (general factors which reduce cellular plasticity).

Here adaptive improvisation refers to reduction of stress-like observables analogous to the above notion of biological stress. New problems, in turn, correspond to internal or external failures which compromise the durability of the system and for which a troubleshooting solution has not been built into the original design. For example, implementation in self-reproducing machines [142] may enable generation of much more complex automata with restorative, replicative, and evolvable capacities. Similarly, implementation in artificial intelligence systems might extend their capacity to solve problems that were not initially taken into consideration.

This principle of self-organization offers a conceptual process enabling adaptation to novel stressful conditions occurring during the lifetime of every single individual. As such, it addresses a critical limitation of natural selection and specifies an efficient new way in which physiology might contribute to evolutionary processes [50, 143, 144]. The combination of restricted variation in potentially deleterious directions and feedback between variation and stress, explains how any type of newly-forming random variation can self-organize into emergent physiological adaptations. Since some of the adaptive changes are also heritable, this proposal principle provides a robust path towards highly broad Lamarckian-like adaptations which are not exclusively dependent on a specific choice of pre-evolved mechanism. The connection between emergent physiological adaptations and progressive improvements within few generations further imply that evolution should be regarded as occurring on every temporal scale and is not limited to changes in allele frequencies.

The proposed picture implicates stress not only in driving newly-forming variation, but also in constantly organizing variation into adaptive outcomes. The above notion of stress is, however, substantially different from the view of stress as activation of a defined stress pathway. Here, we regard stress as a collection of driving forces which operate on diverse processes across multiple scales (from the scale of intracellular processes, through tissues and organs, and up to the whole organism level and above). Regarding stress in this way: 1) accounts for a tendency to re-achieve broad stability (drive reduction), and 2) explains how de novo coordination between emergent adaptations in cells, tissues and organs can arise by parallel stress-reductions in multiple scales.

This proposal portrays gene regulation, development and evolution as a synergy between outcomes of natural selection and ongoing adaptation by improvisation. The latter may be essential for viability, durability and evolvability, and it is possible that no organism can exist without it. The extent of adaptive improvisation increases with the magnitude of stress that was not alleviated by pre-evolved mechanisms. This provides a natural ‘division-of-workload’ between pre-evolved process and adaptive improvisation, without a need for a computational-like ‘decision making’ process. Similarly to natural selection, adaptation by improvisation can involve any kind of mechanism. The principle of adaptive improvisation therefore extends adaptation by natural selection to temporal and organizational scales within single individuals.

Adaptive improvisation can explain why dysregulation leads to formation of heterogeneous tumors which detach from normal coordination and become extremely resistant to drugs. It suggests a modified strategy of cancer therapy which would compromise the improvisation capacity of tumor cells by making use of factors which reduce cellular plasticity.