What determines organ size during development and regeneration?

The Vitruvian Man, da Vinci's illustrious drawing, represents with unsurpassed precision the symmetry and proportions found in the human body. For renaissance artists, body proportions evoked beauty and the place of Man at the centre of the universe. For developmental biologists, these properties relate to a complex series of phenomenon bridging genetics, signalling, mechanical feedback, robustness and precision. Mechanistically, the control of organ size results from the complex integration of autonomous programmes, whereby intrinsic signals specify organ identity, patterning and growth properties, and systemic controls adjust organ growth to developmental and environmental cues. But what are the respective roles of these two types of control? What are the contributions of deterministic genetic programmes versus self-organizing mechanisms? And what are the growth drivers and the size rulers? These are among the many outstanding questions in this field. In this short Spotlight article, we discuss recent advances and future challenges for this fascinating discipline of biology, which sits at the cross-road of morphogenesis, metabolism, mechanobiology, tissue regeneration, organoid biology, mathematical modelling and evolutionary biology.

The ability of organs to develop independently of their normal environment was first suggested by Ross Harrison in 1924 (Harrison, 1924). In a series of elegant grafting experiments, Harrison was able to swap anterior limb buds between two salamander species of different size and the spectacular result was that limbs grafted into a different species kept their size of origin (Fig. 1). This indicates the capacity of some organs to follow a genetically encoded growth programme while in a different environment. Heterotypic ‘grafting’ has since been achieved successfully using induced pluripotent stem cells (iPSCs) to grow a mouse pancreas in a rat host unable to make a pancreas (Yamaguchi et al., 2017). This opens up the ethically challenging possibility of using animal hosts to grow human organs (Cyranoski, 2019). Interestingly, in the context of pancreatic replacement, mouse iPSCs injected into a rat blastocyst generate a mouse pancreas that has the size of a rat pancreas. This illustrates the various tissue-specific balances existing between autonomous and systemic size control mechanisms, and sets the stage for future exploration of these regulatory mechanisms.

The autonomous growth control properties of organs can also be explored through their ability to recover after perturbation. Many organs can repair from amputation or injury while they develop, but only a few species can regenerate organs following amputation of that fully developed organ. Studying regeneration has uncovered fundamental properties of organ patterning, leading to the concept of intercalary growth. When two distal segments of a cockroach leg are joined together after removing an intermediate part, the leg regenerates the missing intercalary segment (French et al., 1976). The same happens with salamander limbs (Iten and Bryant, 1976). This fits with the notion of ‘positional information’, which is used during development to dictate the shape and size of an organ (Wolpert, 1969, 1989). A discontinuity in positional information is what generates growth in order to smooth out these disparities. This concept has been largely supported by the discovery of morphogens and their diffusion along different axes of vertebrate and invertebrate tissues. Indeed, the field of morphogen signalling has contributed immensely to our understanding of pattern formation and growth (Briscoe and Small, 2015). However, despite intense studies, the links between morphogens and organ/tissue growth are only partially clarified.

Organ growth appears to be controlled at a global level. Indeed, neither cell proliferation nor cell size suffices to drive growth, as a defect in proliferation can easily be compensated for by an increase in cell size in order to maintain normal organ size (Fankhauser, 1945; Neufeld et al., 1998). Morphogens could offer such global control, because they cover relatively large fields of cells by diffusing from their sites of production. However, the graded distributions of morphogens contrast with the almost homogeneous growth and mitotic activities observed across developing tissues, and this paradox has been difficult to integrate into our understanding of growth control. Over the last two decades, intense research has led to several models exploring the growth-promoting activity of morphogens, mainly focusing on the BMP-like morphogen Decapentaplegic (Dpp) in the Drosophila imaginal wing disc (a larval epithelial structure that undergoes extensive growth and morphogenesis to form an adult wing after metamorphosis). In the ‘gradient scaling’ model (Day and Lawrence, 2000), the graded activity of Dpp adjusts to the size of the disc, but Dpp production at the source remains constant. This model proposes that a cell in the gradient activates growth and proliferation when it reads a difference in Dpp activity in its neighbours. As the disc expands, the slope of the gradient is reduced. This explains both the homogenous pattern of cell divisions and the progressive arrest of cell proliferation, which occurs when the difference in Dpp signalling between two adjacent cells drops below a threshold (Rogulja and Irvine, 2005; Wartlick et al., 2011). An alternative ‘temporal dynamics’ model was later proposed, based on the observation that Dpp concentrations increase with time at the source (Wartlick et al., 2011). Here, each cell is exposed to a relative increase in Dpp signalling over time, which promotes cell division. This relative increase declines progressively with time until proliferation ceases. However, whether the Dpp gradient scales with tissue size is still controversial (Hufnagel et al., 2007) and recent studies indicate that the graded nature of Dpp activity, although crucial for patterning, is dispensable for tissue growth (Barrio and Milán, 2017; Bosch et al., 2017).

But how do morphogens promote growth? It is known that cells grow by increasing their protein, lipid and nucleic acid contents. One possibility could be, therefore, that morphogens simply allow an underlying basal growth machinery to operate. Indeed, Dpp acts on growth by repressing the transcription factor Brinker (Brk), which represses the growth activators Myc and Bantam in lateral cells. Co-expression of brk, bantam and Myc in the wing pouch only partially rescues the growth inhibition mediated by brk alone, suggesting that Brk could act on other targets for growth inhibition (Doumpas et al., 2013). Similarly, the morphogen Wingless (Wg) promotes growth by counteracting the repressor form of the transcription factor TCF (also known as Pangolin in Drosophila). When a morphogen, either Dpp or Wg, and its negative effector are simultaneously inactivated, wing imaginal discs grow almost normally (Barrio and Milán, 2017, 2020; Schwank et al., 2008). This suggests that Dpp and Wg act by preventing growth inhibition by Brk and TCF. This forms the basis of the ‘growth equalization model’ whereby Dpp serves to balance the non-uniform growth potential of wing disc cells by restricting the growth repressor Brk to the lateral domain (Schwank et al., 2008). How Dpp and Wg function independently of one another to drive growth remains unclear. Their respective overexpression induces distinct anisotropic growth patterns, suggesting that they carry non-interchangeable growth functions (Barrio and Milán, 2020).

The paradigms emerging from studies of patterning and growth in Drosophila have also been interrogated in vertebrates, which exhibit an increased range of organ size. These studies have revealed that, in structures such as the limbs or the spinal cord, early cell-type specification is dictated by gradients of morphogens (e.g. BMPs, Wnts, sonic hedgehog). This early specification phase precedes a growth phase (Kicheva and Briscoe, 2015). This allows larger structures to form without being limited by the short-range effects of morphogen diffusion. In this context, however, maintaining organ proportions during the growth phase requires regulative feedback strategies to correct for variability in the number of precursor cells (Lander et al., 2009).

In summary, despite intense research, our understanding of the role of morphogens in tissue growth is far from complete. This is due to the inherent difficulty of observing morphogen actions in vivo, which moving forward will necessitate the use of more-powerful, less-disruptive technologies. These include refined gene-editing techniques to manipulate endogenous gene function, non-disruptive live imaging to evaluate the kinetics of endogenous morphogens, and fast light-mediated gene activation/repression to avoid long-term adaptation effects due to genetic manipulation.

As highlighted above, organ patterning and growth rely on autonomous signals. However, the picture is much more complex than this. Using inter-species eye-grafting experiments in salamanders, similar to the limb-grafting work of Harrison, Victor Twitty realized over 90 years ago that grafted eyes did not grow on starved hosts, and that the ability of the graft to grow decreased with age (Twitty, 1930). This led Twitty to conclude that ‘the potential size of (an organ) is largely determined by intrinsic factors, but that the expression or realization of this potentiality during the growth of the animal depends upon its interactions with gradually changing environment’. Many would still agree with this conclusion today. Indeed, it is now known that growth strongly relies on environmental factors. These include light, humidity, temperature, crowding, oxygen and nutrients; the importance of the last of these factors – nutrition – was tragically illustrated by studies linking nutrient restriction to a decrease in the size of newborns and children during World War II (Angell-Andersen et al., 2004; Stein et al., 2004).

The control of organ and body growth by nutrition involves sensor and effector pathways. Since the discovery of the Target of Rapamycin (TOR) network in yeast and its conservation in metazoans, we have gained a better knowledge of the cellular machinery that couples nutrients with cell and tissue growth (Liu and Sabatini, 2020). At the heart of this machinery lies the TORC1 complex, which senses energy and nutrient inputs, and controls cell growth autonomously through the coordinated activation of protein, lipid and nucleotide synthesis. In Drosophila, hypomorphic Tor mutations produce viable small animals, illustrating the essential role of this pathway in the general control of cell, organ and body size (Zhang et al., 2006). TORC1 also receives inputs from the insulin/insulin-like growth factor (IGF) signalling pathway, which coordinates nutrient availability and growth through circulating hormones. Indeed, a coordinated reduction in all (or most) body parts is observed under nutritional restriction, highlighting that systemic hormonal relays play a key role in adjusting organ and body growth in response to nutrient availability. Outstanding questions in the field therefore have focussed on the identity of nutrient sensor tissues and the nature of the nutrients sensed, the identity of the signals ensuring organ crosstalk for systemic growth regulation, and the definition of the effector pathways.

The power of genetic analyses in Drosophila facilitated the study of nutritional physiology. Specifically, the possibility of manipulating different genetic pathways in separate organs allowed the identification of a large number of physiological crosstalk events involved in nutritional response (Droujinine and Perrimon, 2016). The fat body, which is functionally related to the vertebrate liver and fat, is key for nutrient sensing in Drosophila. Fat body cells use TORC1 and possibly other branches of the nutrient-responding pathways to: (1) store/release nutrients; and (2) produce and send signals triggering a systemic response in peripheral tissues. Fat body-derived signals activate specific centres in the brain, where Drosophila insulin-like peptides (Dilps) are secreted to promote systemic growth when conditions are favourable (Agrawal et al., 2016; Delanoue et al., 2016; Koyama and Mirth, 2016; Meschi et al., 2019; Rajan and Perrimon, 2012; Sano et al., 2015). Muscle tissue (Katewa et al., 2012) and the gut (Alfa et al., 2015; Redhai et al., 2020; Rodenfels et al., 2014; Song et al., 2017) are also emerging as major nutrient sensors. Interestingly, many factors identified in cross-organ communication in Drosophila are also found in vertebrates (e.g. Dilps/insulin/IGF, Impl2/IGF-BP, AKH/glucagon, Upd2/Leptin, Egr/TNFα, Myo/GDF11, Daw/TGFβ, Hh/Shh, GBP/EGF), suggesting functional conservation of their physiological roles. However, our understanding of how these factors come together to achieve systemic growth regulation remains incomplete.

Moving forward in this direction will first require a clear description of nutrient sensors and their respective nutrient stimuli. In invertebrate genetic models, this could reasonably be achieved through exhaustive screening of organs and pathways for systemic growth defects (for recent examples, see Delanoue et al., 2016; Agrawal et al., 2016; Redhai et al., 2020). Moreover, current descriptions of nutrient sensing have thus far been limited to broad categories, i.e. lipids, carbohydrates and amino acids. However, cells need to distinguish between nutrients with high precision. A clear illustration of this is the exquisite regionalization of specific amino acid transporter expression in spatially structured organs such as the gut and the liver (Kandasamy et al., 2018), suggesting that localized nutrient sensing is adjusted to the needs of specific cell types. Our current view of global nutrient sensor functions should thus be revisited to integrate the notion of cell- or region-specific sensors, as recently exemplified in the case of copper-sensing cells in the fly gut (Redhai et al., 2020).

Another future challenge is to understand the central integration of nutritional signals. In both mouse and Drosophila models, recent findings indicate that a major part of this integration takes place in the brain. Nutrients enter the brain and directly activate neural sensors controlling energy balance and feeding (Augustine et al., 2018). Alternatively, the brain receives relay signals produced by nutrient-sensing organs in the form of hormones (e.g. leptin, insulin, adiponectin and several gut-derived peptides), which participate in the central control of energy balance (Kim et al., 2018). This has led to an integrated model in which the vertebrate hypothalamus, together with the central nervous system, forms a complex neural network controlling energy homeostasis. Our ability to fully decipher this circuitry controlling nutritional homeostasis is, of course, challenged by the complexity of the mammalian brain. Important insights could therefore come from parallel studies on simpler brains, such as those of insects for which the elucidation of full neural architecture is now within reach (https://www.janelia.org/project-team/flyem). In addition, recent improvements in optogenetic and chemogenetic techniques could allow precise manipulations of nutrient-sensing centres.

An additional and important aspect of nutritional physiology that has emerged recently is the observation that not all body parts are equally sensitive to nutrient restriction during growth. Indeed, a hierarchy of tissues is established in response to nutrient shortage, as illustrated by early studies on intrauterine growth restriction. In infants suffering nutrient deprivation, the redistribution of blood flow maximizes oxygen and nutrient supply to the brain, allowing this specific organ to be spared (Cohen et al., 2015). Remarkably, this feature is also observed during development in mice (Gonzalez et al., 2016) and flies (Cheng et al., 2011). In Drosophila, the observation of phenotypic plasticity in response to nutrition identified another hierarchy among imaginal discs. Whereas genital discs appear rather insensitive to nutritional deprivation, others are strongly affected, leading to various degrees of size decrease and altered body proportions. Molecularly, this hierarchy relies on the ability of different organs to ‘read’ nutritional information through insulin signalling (Tang et al., 2011). This finding therefore highlights that the proportions between different organs can be modified through a plastic response to changing environments.

What happens when the growth of one body organ is impaired provides interesting insights into a more global level of regulation taking place between body parts. In the early 1980s, it was observed that abnormal growth of Drosophila imaginal discs (triggered by injury, undergrowth or tumoural growth) delays developmental transitions (Simpson et al., 1980). These inspiring findings indicated that ill-growing tissues can modify body physiology through systemic signals. A relaxin-like hormone called Dilp8 has since been found to be produced by growth-impaired discs. Dilp8 delays pupal development by inhibiting production of the steroid hormone ecdysone through a central brain relay (Colombani et al., 2012, 2015; Garelli et al., 2012, 2015; Vallejo et al., 2015). In addition, for body proportions to be maintained when the growth of one organ is impaired, the growth rates of other body parts must adjust. Such coordination is observed between imaginal discs (Parker and Shingleton, 2011), or between regions of a single disc (Gokhale et al., 2016; Mesquita et al., 2010). In both cases, this coordination relies on the action of Dilp8 on ecdysone levels (Boulan et al., 2019; Sanchez et al., 2019). Therefore, the systemic regulation of both developmental time and growth rate relies on a single molecular mechanism, allowing tight coordination between developmental parameters (Fig. 2A). Despite important progress, many aspects of this coordination remain. In particular, it is not known how systemic relays act differentially on the injured parts (which compensate and grow faster) and the healthy ones (which slow down and ‘wait’).

In vertebrates, long bones are the model of choice for studying limb growth regulation, growth termination and inter-organ communication (Roselló-Díez and Joyner, 2015). Here again, new tools have extended the power of genetic manipulations and have paved the way for novel findings. In the mouse model, a complex genetic set up has been used to allow for unilateral inhibition of cell proliferation in the growth plate of the left limb (Roselló-Díez et al., 2018). This approach revealed that contralateral growth coordination takes place during early pup development (Fig. 2B), although the molecular mechanisms mediating this crosstalk are unknown.

Contralateral limb communication also appears to play a role in the context of regeneration. Early experiments on salamander and axolotl have shown that, after amputating two limbs on the same animal, regeneration progresses slower when the second amputated limb is contralateral to the first one, possibly because of transneuronal connexions (Tweedle, 1971). In Xenopus, unilateral leg amputation triggers a mirrored bioelectric signal on the contralateral limb, suggesting that depolarization could act as a long-range signalling process (Busse et al., 2018). Moving forward, merging the fields of growth control and limb regeneration could be instrumental, allowing us to revisit the early findings derived from grafting and regeneration experiments and provide further insight into inter-organ communication.

Growth stops when organs or tissues reach their target size. However, behind this simple statement lies some very complex biology. As presented earlier, appendage grafting and transplantation experiments in salamanders and Drosophila suggest a ‘ruler’ mechanism that is intrinsic to each organ. In fact, different organs use different ways to assess their target size. The determination of liver size, for example, depends on functional feedback that follows a liver:body size ratio. Accordingly, if transplanted into a host with a different size from the donor, the liver adjusts to the size of the host (Kam et al., 1987). The same applies to the thyroid gland, which increases its size when thyroid hormone production is deficient due to a negative-feedback loop involving thyroid-stimulating hormone (Ortiga-Carvalho et al., 2016). These are examples of ‘regulative’ control, whereby final organ size relies on extrinsic cues, mostly linked to function. By contrast, the gut, the pancreas or the thymus do not adjust to recipient size after transplantation, nor do they regrow to normal size after ablation early in development (Metcalf, 1964, 1963; Stanger et al., 2007). In the case of the pancreas, the number of divisions for each progenitor cell appear to be limiting, therefore preventing full regeneration upon ablation (Stanger et al., 2007).

Organ mass can also be used as a ruler for final organ size. In this case, feedback factors called ‘chalones’ have been proposed to be secreted proportionally to organ mass and inhibit growth in an autocrine manner. Although this concept is attractive, only a few experimentally confirmed examples of chalones have been described to date. These include the TGFβ myostatin in skeletal and cardiac muscles (McPherron et al., 1997) and GDF11 in the olfactory neural epithelium (Gamer et al., 2003). In the vertebrate limb bud, the accumulation of tissue mass displaces signalling centres and, as a consequence, blocks a positive-feedback loop between the morphogens Shh and Fgf, inducing proliferation arrest (Scherz et al., 2004; Verheyden and Sun, 2008). As discussed above, morphogens are essential for patterning and growth during development. Two separate models, the gradient scaling model and the temporal dynamics model, propose that morphogen signalling also contributes to growth arrest and the determination of final organ size in the fly. Although supported by elegant experiments, these models are challenged by contradicting experimental evidence (Vollmer et al., 2017). In the gradient scaling model, proliferation should stop when the Dpp signal is uniform. Yet, wing growth is observed when Dpp activity is constant throughout the disc (Bosch et al., 2017; Schwank et al., 2008). In the temporal dynamics model, a relative increase of Dpp signalling is needed to sustain proliferation. However, clones of cells lacking Mad (the intracellular mediator of Dpp) and the growth inhibitor Brk grow to a normal size despite being unable to transduce an increase in Dpp signal (Schwank et al., 2012).

Mechanical forces exerted on cells within tissues form an additional level of regulation. At the tissue scale, cell proliferation and growth increase local pressure, inducing mechanical feedback that inhibits growth in compressed zones and stimulates growth in stretched ones. Two distinct mathematical models integrating both morphogen signalling and mechanical feedback predict uniform proliferation and growth termination in tissues (Aegerter-Wilmsen et al., 2007, 2012; Hufnagel et al., 2007; Shraiman, 2005). In these models, growth stops when compression eventually overcomes the non-uniform growth-promoting activity of morphogens. Experimental measurements comparing patterns of proliferation, mechanical stress and cell deformation in the central and lateral regions of the Drosophila wing disc support this model (LeGoff et al., 2013; Mao et al., 2013; Nienhaus et al., 2009). Mechanistically, this feedback involves the regulation of cell proliferation and apoptosis via changes in the acto-myosin cytoskeleton and the canonical effectors of the growth-promoting Hippo/Yorkie (Hpo/Yki) pathway (Pan et al., 2016; Rauskolb et al., 2014). However, a recent report showing that releasing tension in the wing disc does not affect Hippo signalling or wing size (Ma et al., 2017) challenges the role of mechanical feedback in controlling final appendage size and leaves the door open for further studies. Although the Hippo effectors Yap/Taz were first shown to transduce mechanical information in vertebrate cells (Dupont et al., 2011), a role for this coupling in controlling organ size in vertebrates has not yet been established.

Regardless of its underlying mechanism, growth termination seems to be controlled organ autonomously. However, final organ size is largely influenced by systemic cues that superimpose a layer of control on autonomous regulatory circuits. In Drosophila, nutritional status, acting via the PI3K and TOR signalling relays, modulates Yki localization and hence the sensitivity of cells to mechanical feedback and final organ size (Borreguero-Muñoz et al., 2019). Interestingly, when the size of the wing is modified by alterations to insulin signalling, the resulting appendages are perfectly patterned, suggesting that morphogen activity accommodates for varying tissue size (Leevers et al., 1996). Indeed, when PI3K activity is specifically modified in the posterior compartment of the wing disc, the domain of phosphorylation of the transducer MAD and the domain of expression of the Dpp target Spalt adapt to compartment size early during wing disc development (Teleman and Cohen, 2000). Nutrient-responsive pathways also modulate the expression of Dally, a proteoglycan involved in the spreading and activity of Dpp, providing a mechanism for morphogen gradient scaling by nutrition (Ferreira and Milán, 2015). Therefore, in addition to promoting growth, Dpp signalling can mediate modifications in tissue size driven by environmental cues.

Finally, an intriguing aspect of growth arrest comes with the observation that the final size of organisms could be limited by the rate of energy flux towards organs and cells. Pioneering work in the late 1940s indicated that the metabolic rate of many animal species scales with body mass with a power law of ¾ (B∝B₀M^3/4; Kleiber, 1947). Remarkably, this so-called ‘Kleiber's law’ applies to organisms as diverse as mammals, birds, fish, insects, molluscs and plants, and for masses ranging across more than 16 orders of magnitude (see Box 1). The theoretical modelling of this empirical finding has remained elusive. Recently, an attractive model was proposed to explain the so called ‘¾ rule’ relying in part on the physical properties of branched, fractal networks that are required to equally allocate metabolic energy to all cells of an organism (West et al., 1997). An important consequence of Kleiber's law is that, whereas energy needs (proportional to cell number) scale to the power of 1 with mass, energy allocation scales to the power of ¾. This imbalance is expected to ultimately limit growth to an asymptotic body mass specific for each species. This led West, Brown and Enquist to propose a universal, dimensionless, growth curve on which growth parameters collected from animal species, including shrimp, shrew, salmon, hen, caw and others, all align perfectly (West et al., 2001). Therefore, allometric scaling of metabolic rate and the energy cost of producing/maintaining biomass could explain body growth arrest. Current research in the field of growth control has only begun to explore the physiological basis of Kleiber's law (Thommen et al., 2019). Future research could help to decipher whether similar principles could also be applied to autonomously limit the growth of individual organs.

The precision of organ size determination is particularly visible for bilateral organs, which in many cases differ in size by less than 1%. These observations raise outstanding questions with regard to the mechanisms of size adjustment. Bilateral organs follow identical genetic programmes and develop within a shared physiological environment. They are therefore ideal biological models for studying precision and stochastic variations in growth during development. Experimentally, these variations can be quantified by measuring the fluctuating asymmetry (FA) index of bilateral traits within individuals (Juarez-Carreño et al., 2018). FA can be readily assessed for many traits in many species, including humans (Graham and Özener, 2016). However, very little is known about the molecular mechanisms controlling FA, developmental stability and precision. In particular, it is still debated whether developmental precision relies on intrinsic properties of gene regulatory networks controlling developmental processes, or on dedicated molecular mechanisms measuring and correcting errors (Félix and Barkoulas, 2015).

The recent identification of genes involved in developmental precision in genetically tractable models paves the way to addressing these questions (Debat and Peronnet, 2013; Félix and Barkoulas, 2015). In Drosophila, the chaperone Hsp90 (encoded by Hsp83) was first shown to correspond to the definition of a ‘robustness factor’, impairment of which induces low-penetrance developmental abnormalities (Rutherford and Lindquist, 1998). Interestingly, Hsp90 was also proposed to serve as an ‘evolutionary capacitor’, allowing the accumulation of cryptic genetic variations. Overexpression of the transcriptional regulator Cyclin G in wing imaginal discs is also sufficient to induce FA autonomously in adult wings (Dardalhon-Cuménal et al., 2018; Debat et al., 2011). Cyclin G acts through the chromatin modifier complexes PRC1 and PR-DUB to coordinate expression of a large number of genes, supporting the idea that specific gene regulatory networks ensure developmental stability. By contrast, the Dilp8 hormone and its receptor Lgr3 control developmental stability at a systemic level through a central relay, and animals mutant for Dilp8 or Lgr3 accumulate developmental errors, quantified by high wing FA (Boone et al., 2016; Colombani et al., 2015; Garelli et al., 2012, 2015; Vallejo et al., 2015). As mentioned above, the Dilp8-Lgr3 axis also mediates inter-organ communication in response to growth perturbation, suggesting that organ crosstalk could participate in developmental stability (Juarez-Carreño et al., 2018). Specific metabolic adaptations can also buffer the effect of dietary stress on development. For example, trehalose metabolism in Drosophila buffers variations in glycemia and ensures homeostasis upon dietary stress (Matsushita and Nishimura, 2020), potentially linking developmental noise with metabolic processes. Interestingly, flies mutant for the proapoptotic gene hid also show increased wing FA (Neto-Silva et al., 2009), suggesting that cell elimination could be used to fine-tune organ size. In future work, genetic and tissue-specific analyses could help to establish how these different inputs act together to control robustness in a biological system.

To gain insights into the temporal and developmental basis of body symmetry, one would ultimately need to follow bilateral organ growth over time with high precision. This type of analysis is now possible with the help of sophisticated imaging techniques. Such work has shown that, in the zebrafish inner ear, left-right size variability is initially high and progressively resolves to achieve symmetry (Green et al., 2017). Similar experiments performed in the zebrafish otic vesicle reveal that mechanical feedback by lumenal hydraulic pressure is required to adjust organ growth rate (Mosaliganti et al., 2019). Vertebrate limbs are initially subjected to intrinsic and extrinsic symmetry breaking signals (Allard and Tabin, 2009; Grimes, 2019). In some genetic backgrounds, symmetric organs become directionally asymmetric, as observed in Holt–Oram syndrome in humans, where only the left arm exhibits severe developmental defects (Newbury-Ecob et al., 1996). This suggests a model in which symmetry is achieved through mechanisms protecting against asymmetric cues (Grimes, 2019). Whether communication between the two bilateral sides is required for proper size adjustment is not known, but identification of the signal(s) required for contralateral coordination upon growth perturbation could aid our understanding of how developmental precision is achieved at a molecular level in mammals.

The complex and rich phenomenology of growth control has inspired research spanning molecular and physical studies of the cell, to integrated organ physiology and evolutionary perspectives. As such, the challenge for future work will be to integrate this multi-scale information into a unified conceptual framework. The various aspects of size control presented in this Spotlight define two distinct visions on how size is coded in biological systems. Genetic and patterning aspects suggest a deterministic mode of control, whereby initial parameters encoded in the early embryo determine the final size of future organs and organisms. This gene-centric view of growth processes is supported in the fly model by the role of morphogen signalling networks in growth and patterning. However, accumulating evidence for feedback mechanisms involving physical or metabolic inputs supports the existence of a self-determined mode of organ growth/size control. Future research will no doubt help to integrate these two ideas. This research should also focus on system energetics and push forward our understanding of the physiological implications of Kleiber's law, in particular at the organ level. Finally, the study of non-physiological conditions, such as growth impairment, regeneration or uncontrolled growth, will be needed to help shed new light on systemic or cross-organ regulatory mechanisms that are, at present, poorly understood.

We thank Yohanns Bellaiche and Paula Santabárbara-Ruiz for useful comments on the manuscript, and Bertsy Goic for drawing the figures.