Biological Parsimony and Interactomes I
In previous posts (April 2015; August 2015) we have looked at the notion of biological parsimony from several vantage points. For example, one such issue was the frequency of certain protein folds as recurring evolutionary motifs, in contrast to other folds which are used much more restrictedly (April 2015 post). Here we look at parsimony at the higher level of biological systems, chiefly concerning the strong tendency for such systems to evolve towards strongly economic arrangements.
Two Levels of Parsimony
In the post of April 2015, a number of different forms of biological modularity were listed, as applied towards parsimonious biosystems. Here we can ‘parsimoniously package’ the general phenomenon of biological parsimony itself into two major levels: that of the packing or arrangement of function in specific molecules or encoded biological information, and that of the deployment of molecules in terms of their functional interactions in the operations of biosystems. Of course, these are not independent factors, since a polyfunctional protein (for example) will have multiple types of interactions with other functional partners within the biosystem within which it operates. This is one means whereby a single protein can have distinct roles in cells of divergent differentiation lineages.
These levels of parsimony and their interactions are depicted in Fig. 1.
Fig. 1. Two major levels of macromolecular biological parsimony and their inter-relationships, schematically depicted. A, Packing / Functional Arrangement refers to parsimonious packaging of encoded information (such as a single genetic coding locus capable of producing multiple distinct proteins, via alternate promoters, differential splicing, or other means; represented here as ‘Informational Encoding’). Also, encompassed within this level is the parsimonious use of encoded macromolecular structures by evolutionary selection (“Evolutionary motif redeployment”) with divergence of function, well-exemplified by the TIM-barrell protein motif, as mentioned in a previous post (April 2015). In addition, the ‘packing’ levels includes the grouping of multiple distinct functions into a single macromolecule (such as a protein W with n separable functions). B, Parsimony at the level of functional interactions. These include intermolecular interactions (for example, where a protein W interactions with n different partner molecules with n distinct results), and also intramolecular effects (as in the case of allosteric changes in a protein induced by ligand binding at a distinct site). Also within this level of parsimony is the evolutionary redeployment of portions of specific signaling pathways in different cellular contexts, with divergent ‘read-out’ consequences.
Aside from the above ‘packing’ issues, another aspect of molecular functional packing which is relevant to biological parsimony at the level of individual molecules is evolutionary ‘redeployment’ parsimony. (In this respect, see also a previous post [April 2015], where this has also been discussed). Such an evolutionary form of parsimony refers to the tendency of biological components, especially at the molecular structural level, to be ‘repurposed’ by evolution towards assuming new functional roles. (This facet of parsimony is also noted in Fig. 1 above.) Why should this be so? In fact, it follows fairly simply from the principle that natural selection process can only ‘tinker’ with what is currently available, and cannot foresee the optimal solutions to biosystem problems which might become apparent in hindsight. Thus, it is usually more probable that encoded pre-existing structures will be co-opted for other functions than wholly new genes will arise de novo. In order for the ‘re-purposing’ to happen, an obvious problem would seem to arise from the simple question, “if some cellular mediator assumes a new function, what takes care of its original function?” In fact, there are well-known processes whereby this can happen, primarily involving the generation of additional gene copies (gene duplication events) with which evolution can ‘tinker’, without compromising the function of the original gene product.
Of course, it could be argued that since the entirety of biology is evolutionarily derived, that all biological parsimony is ‘evolutionary’. In a broad sense, this is obviously true, but it is worth highlighting the ‘repurposing’ type of evolutionary parsimony for special mention in this context. It is certainly true that not all evolutionary change can be classified as arising from the redeployment of pre-existing ‘parts’ towards novel applications, in any case. For example, where a single mutation in a functional protein confers a selectable fitness benefit, which ultimately becomes fixed in a population, evolutionary change has occurred – but via a direct modification of an existing ‘part’ towards better efficiency in its original role, not towards an entirely new function.
That matter aside, in this parsimonious post, the focus will be on interactomes, following a brief introduction from the previous post on this topic.
It was noted in the previous post that the seemingly low numbers of coding sequences in human and other ‘higher’ organisms is counter-balanced to a considerable degree by various diversity generating mechanisms, by which a single gene can encode multiple proteins, or a single protein can be post-translationally modified in distinct ways. But as well, it was also noted in passing that many (if not most) proteins have more than one role to play in developing organisms, often in cell types at distinct stages of differentiation. This is the essence of the interactome, the sum total of molecular interactions that enable a biosystem to function normally. In this context, a key word is connectivity, where ‘no protein [or any functional mediator] is an island, entire to itself’.
There are numerous ways that the parsimony principle is manifested within interactomes. One prominent feature in this regard is signaling and signaling pathways. It is common to find a single defined signaling mediator with multiple roles towards different cell types, or at different stages of differentiation. An example to consider here is a cytokine known as Leukemia Inhibitory Factor, or LIF. As its name suggests, it was first defined as a factor inhibiting the growth of leukemic cells, yet in other circumstances it can behave as an oncogene. It is well-known as a useful reagent in cell biology owing to its ability to maintain the pluripotent differentiation status of embryonal stem cells, an activity of great utility for the generation of ‘knock out’ mice. But in addition to this, LIF has been shown to have roles in the biology of blastocyst implantation, hippocampal and olfactory receptor neuronal development, platelet formation, proliferation of certain hematopoietic cells, bone formation, adipocyte lipid transport, production of adrenocorticotropic hormone, neuronal development and survival, muscle satellite cell proliferation, and some aspects of hepatocyte regulation. An irony of this polyfaceted range of functions is that certain activities among the above LIF-list were at first ascribed to new and unknown mediators, before detailed biochemical analysis showed that LIF was the actual causative factor.
The extent of the pleiotropism (‘many turns’) of LIF has intrigued and surprised numerous workers, leading to this effect being called an ‘enigma’. Why should one cytokine do so many things? Here it should be noted that in the cytokine world, while LIF is certainly not unique in having multiple activities, it is probably the star performer in this regard. In answer to the question “why does it make design sense to use LIF in the regulation of such a diverse and unrelated series of biological processes?”, we can invoke the parsimony principle, by a now familiar logical pathway. It is thus reasoned that a biosystem factor will tend to assume multiple functional roles if it can do so without compromising organismal fitness. The ‘tend to’ phrase is predicted on the assumption that is energetically and informationally favorable to streamline the functional operations of a biosystem as much as possible, and that evolutionary processes will move organism design in that direction, via increased fitness gains. At the same, it is evident that there must be limits to this kind of trend, since at some point in the poly-deployment of a mediator, inefficiencies will inevitably creep in, as one signal event begins to interfere with another. A number of ways have been ‘designed’ by evolution to minimize this, of which more below. But to return to the specific question of why should LIF – and not some other cytokine – be such an exemplar of polyfunctionality, there is no specific answer. All that can be suggested is that the many biological roles that feature LIF do not interfere with each other, or that they complement each other, such that there is a fitness gain by LIF’s multideployment in such ways. And this could be condensed into saying, ‘it can, so it does’, which might not sound particularly helpful. There may be reasons of simple evolutionary contingency as to why LIF gained these roles and not some other cytokine – or indeed there may be deeper (and highly non-obvious) reasons why the prize necessarily must go to LIF. Such questions might be answered at least in part by advanced systems biological modeling, or (ultimately) by equally advanced synthetic biology, where artificial programming of real-world model biosystems can address such questions directly.
With this introduction in the form of LIF in mind, it is useful to now think about ways that receptor signaling can diversify with either a single mediator involved, or with a single receptor. With respect to the latter circumstances, there are biological precedents where a single heterodimeric receptor (composed of two chains) can respond with distinct signaling resulting from engagement with separate ligands. This effect is well-exemplified by the Type I interferons (IFN), of which there are several distinct types (in humans alone, these include IFN-α, IFN-β, IFN-ε, IFN-κ, and IFN-ω, where IFN-α has 13 different subtypes), all of which bind to the same heterodimeric Type I receptor. Yet despite their sharing of a common receptor, the signaling induced by these distinct kinds of interferons is quite distinct as well. This phenomenon is depicted in Fig. 2 below.
Fig. 2. Schematic depiction of a single heterodimeric receptor which enables distinct signaling from binding of different ligands, even in the same cell type. In the top panel, Ligand A (blue pentagon) engages certain specific residues within the receptor pocket, with induction of a conformational change which activates a subset of the intracytoplasmic co-signaling molecules, with a specific signaling pathway triggered. The bottom panel depicts a different ligand (Ligand B, red hexagon), which engages the receptor with different contact residues, resulting in distinct receptor changes and concomitant downstream signaling.
In general, the form of ligand signaling complexity depicted in Fig. 2, where a specific ligand can activate one signaling pathway without activating another, has been termed ‘biased agonism’. This phenomenon has been much-studied in recent times with respect to G-Protein Coupled Receptors (GPCRs), which are a hugely diverse class of cellular receptors. They have long been of particular interest to the pharmaceutical industry through their susceptibility to selective drug action (‘druggability’), and biased agonism clearly offers a handle on improving the selectivity by which GPCR-mediated signaling is directed in a desired manner.
Other complexities to signaling arrangements are possible which increase signal diversity from a limited set of participants. Cells of different lineages may express the same receptors, but differ in their patterns of co-receptors and signaling accessory molecules whereby intracellular signals are generated. This is depicted in Fig. 3A and Fig. 3B below. Other processes whereby a limited set of ligands and receptors diversify their signaling are shown in Fig. 3C- Fig. 3F. Thus, signaling-based polyfunctionality is one aspect of interactomic parsimony.
Fig. 3. Schematic depiction of mechanisms for signaling diversity generated with either the same receptor in different contexts (A-E), or the same ligand binding to a different receptor (F). A and B: the same receptor (as a heterodimer) expressed in cells of two distinct differentiation states, such that they differ in their complements of coreceptors (not shown) or intracytoplasmic accessory signaling molecules (colored ovals). After engagement with ligand, the resulting signal pathway in context A is thus divergent from that generated in context B; C and D: the same receptor where it forms a homodimer (C) or heterodimer (D), each with distinct signaling consequences; E: the same receptor as in A, but where it interacts with a second ligand (pale blue octagon), which engenders a conformational change such that it binds either a different ligand, or a modified form of the original ligand; F: the same ligand as in A, but where it is compatible with another receptor entirely, with corresponding divergent signaling effects.
The deployment of different subunits in the signaling arrangements of Fig. 3 is itself a subset of a more general effect within interactomes, where modularity of subunits within protein complexes is a ubiquitous feature. This reflects an aphorism coined in a previous post (April 2015), to the effect that “Parsimony is enabled by Modularity; Modularity is the partner of Parsimony”. And with respect to protein modularity in eukaryotic cells, there is plenty of evidence for this from studies of the yeast proteome, where differential protein-protein combinations have been extensively documented.
Signaling to different compartments
Biosystems are compartmentalized at multiple levels. As well as the unit of compartmentalization we know as cells, numerous membrane-bound structures are ubiquitously encompassed within cellular boundaries themselves. An obvious one to note is the cell nucleus itself. While the subcellular organelles known as mitochondria (the energy powerhouses of cells) and chloroplasts (the photosynthetic factories of green plants) have their own small genomes encoding a limited number of proteins, in both cases many more proteins required for their functions are encoded by the much larger host cell genomes. Other compartments lacking their own genomes exist, including (but limited to) the endoplasmic reticulum, the Golgi apparatus, and peroxisomes.
It would be easy to imagine a host genome-encoded set of special proteins reserved for the organelles or other compartments, along with specialized transport systems (to target the organelle-required proteins to the right places) in each case. In some cases, this appears to be so, but if this was generalized, it would certainly violate the parsimony principle, since many such proteins are also required to function in more than one cellular compartment. One could envisage a solution in the form:
Signal A – Protein 1, 2, 3….. | Signal A recognition system, to compartment A
Signal B – Protein 1, 2, 3….. | Signal B recognition system, to compartment B
By such an arrangement, an identical set of proteins could be targeted to distinct compartments if they were are appended to modular recognition signals. Yet as is so often the case, biology is both more subtle and more complicated than simplistic schemes such as this. In fact, a variety of natural ‘solutions’ for the multi-targeting issue have evolved. To use the above terminology, some could be depicted at the mRNA level as:
Signal A (spliced in) – Protein 1 coding sequence….. | (expression) — Signal A recognition system, to compartment A
Signal A (spliced out) – Protein 1 coding sequence….. | (expression) — no targeting signal, remains in cytosol.
In these circumstances, the ‘default’ localization is with the cytoplasm (cytosol), and organelle targeting is effected only where a signal sequence is translated and appended to the protein. Differential splicing at the RNA level can then include or exclude the sequence of interest, both (parsimoniously) from the same genetic locus. But many more mechanisms than this have been documented for general multi-compartmentalization, including the existence of chimeric signal sequences that are bipartite towards different compartments. The take-home message once again is the stunning extent to which known biosystems have evolved highly parsimonious deployment of their encoded functional elements, all encompassed within biological interactomes.
This short tour of the parsimonious interactome has barely scratched the surface of the topic as a whole, and some other aspects of biology parsimony will indeed be taken up in the next post. Meanwhile, a biopoly(verse) take on receptor-ligand parsimony:
A ligand-receptor attraction
Can show parsimonious action
For receptors can change
In their signaling range
And vary a transduced reaction
References & Details
(In order of citation, giving some key references where appropriate, but not an exhaustive coverage of the literature).
‘…..it is usually more probable that pre-existing structures can be co-opted for other functions than wholly de novo genes will arise.’ It has long been considered that gene duplication is an effective means by which novel functions can evolutionarily arise, and far more likely than de novo gene evolution. In this regard, see a review by Hurles 2004. Yet in recent times evidence for the evolution of de novo genes from ‘orphan’ open reading frames has become stronger; see Andersson et al. 2015. Nevertheless, the duplication-mediated repurposing of pre-existing evolutionary ‘parts’ is still most likely to be much more frequent.
‘….Leukemia Inhibitory Factor, or LIF. As its name suggests, it was first defined as a factor inhibiting the growth of leukemic cells….’ For general LIF background and its polyfunctional nature, see Hilton 1992 and Metcalf 2003. As other examples of LIF anti-tumor activities, see Bay et al. 2011; Starenki et al. 2013.
‘…..yet in other circumstances it [LIF] can behave as an oncogene.‘ See Liu et al. 2015.
‘……this effect [LIF polyfunctionality] being called an ‘enigma’ …..’ See Metcalf 2003.
‘…..“why does it make design sense to use LIF in the regulation of such a diverse and unrelated series of biological processes” ……’ This question (slightly paraphrased here) was posed by Metcalf 2003.
‘……This effect is well-exemplified by the Type I interferons.’ See a review by Platanias 2005.
‘…….‘biased agonism’. This phenomenon has been much-studied in recent times with respect to G-Protein Coupled Receptors (GPCRs)…..’ For very recent updates on biased agonism in a GPCR context, see Pupo et al. 2016; Rankovic et al. 2016.
‘……protein modularity in eukaryotic cells, there is plenty of evidence from studies of the yeast proteome, where differential protein-protein combinations have been extensively documented.‘ See Gavin et al. 2006; Gagneur et al. 2006.
‘……But many more mechanisms than this have been documented for general multi-compartmentalization.’ See a review by Yogev and Pines 2011, where at least 8 different targeting mechanisms were listed for mitochondria alone. See also Avadhani et al. 2011 for a discussion of chimeric signals in a specific protein context.
Next Post: March.