Encapsulation of ribozymes inside model protocells leads to faster evolutionary adaptation

Functional biomolecules, such as RNA, encapsulated inside a protocellular membrane are believed to have comprised a very early, critical stage in the evolution of life, since membrane vesicles allow selective permeability and create a unit of selection enabling cooperative phenotypes. The biophysical environment inside a protocell would differ fundamentally from bulk solution due to the microscopic confinement. However, the effect of the encapsulated environment on ribozyme evolution has not been previously studied experimentally. Here, we examine the effect of encapsulation inside model protocells on the self-aminoacylation activity of tens of thousands of RNA sequences using a high-throughput sequencing assay. We find that encapsulation of these ribozymes generally increases their activity, giving encapsulated sequences an advantage over nonencapsulated sequences in an amphiphile-rich environment. In addition, highly active ribozymes benefit disproportionately more from encapsulation. The asymmetry in fitness gain broadens the distribution of fitness in the system. Consistent with Fisher’s fundamental theorem of natural selection, encapsulation therefore leads to faster adaptation when the RNAs are encapsulated inside a protocell during in vitro selection. Thus, protocells would not only provide a compartmentalization function but also promote activity and evolutionary adaptation during the origin of life.

RNA is believed to have been a central constituent of early life (1–3). In the “RNA world” theory, functional RNAs (e.g., ribozymes) would both perform catalytic functions and store and transfer genetic information in a simple living system (4–6). Encapsulation of ribozymes in cell-like compartments, such as protocells, is thought to be an essential feature for the emergence of early life (7–11). In particular, compartmentalization would retain useful metabolites in the vicinity (12) and prevent a cooperative, self-replicating ribozyme system from collapsing under parasitization by selfish RNAs (13, 14). A major model of protocells is lipid vesicles, which consist of an aqueous interior surrounded by a semipermeable membrane (15, 16). However, while the ultimate advantages of compartmentalization may be clear, how encapsulation and confinement inside protocell vesicles would affect the activity and early evolution of ribozymes is not understood well.Confinement by lipid membranes presents a biophysical environment similar to macromolecular crowding (17). The effect of macromolecular crowding on the activity, function, and specificity of biomolecules (i.e., proteins and nucleic acids) has been examined extensively (18–23) using crowding agents such as dextran, polyethylene glycol, and Ficoll in vitro (24–29). In general, macromolecular crowding agents decrease the accessible volume for biomolecules, leading to the excluded-volume effect, in which the relative stability of compacted and folded structures is increased (30, 31). At the same time, chemical interactions between the crowding agents and the biomolecule can also stabilize or destabilize the folded structure, influencing catalytic activity (24, 32). While chemical interactions depend on the properties of the specific molecules under study, the excluded-volume effect resulting from spatial confinement inside vesicles is expected to be general. The effect of confinement can be studied while controlling for chemical interactions by comparing the encapsulated condition to the nonencapsulated but membrane-exposed condition. This comparison represents the prebiotic scenario in which RNAs would be present in the same milieu as lipids (33) and may become encapsulated or not. In this way, confinement inside vesicles was shown to increase the binding affinity of the malachite green RNA aptamer (34). Interestingly, spatial confinement inside a tetrahedral DNA framework has also been shown to increase thermodynamic stability and binding affinity of aptamers by facilitating folding (35).While these and other case studies (17, 25, 36–43) illustrate mechanisms by which RNA activity might be perturbed inside vesicles, understanding how encapsulation would affect evolution requires a broader scale of information. In particular, detailed knowledge of how encapsulation affects the sequence-activity relationship is required. This information is captured in the “fitness landscape,” or the function of fitness over sequence space, which embodies many important evolutionary features [e.g., fitness maxima, epistasis, and the viability of evolutionary trajectories (44–47)]. In practice, the fitness of a ribozyme can be considered to be its chemical activity for a particular function in the given environment (48–53).In the present work, we investigated how encapsulation inside model protocells would affect the catalytic activity and evolution of self-aminoacylating ribozymes. We studied tens of thousands of RNA sequences derived from five previously selected self-aminoacylating ribozyme families (53). These sequences were encapsulated in a mixed fatty acid/phospholipid vesicle system. Fatty acids mixed with phospholipids (1:1 molar ratio) have been used as model protocell membranes, as the vesicles tolerate Mg²⁺ concentrations needed for ribozyme activity and the membrane allows small, charged molecules to permeate while preserving large polynucleotides in the vesicle interior (54, 55). To study the biophysical effect of confinement rather than chemical interactions with the membrane, RNA activity inside vesicles was compared with RNA activity when exposed to the same vesicles without encapsulation. We show that ribozymes generally exhibit higher catalytic activity inside the vesicles and that more active sequences experience greater benefit. Using in vitro selection, we demonstrate that one of the evolutionary consequences of this trend is that encapsulation inside vesicles causes a greater rate of genotypic change due to natural selection.     

Niche adaptation and genome expansion in the chlorophyll d-producing cyanobacterium Acaryochloris marina

Acaryochloris marina is a unique cyanobacterium that is able to produce chlorophyll d as its primary photosynthetic pigment and thus efficiently use far-red light for photosynthesis. Acaryochloris species have been isolated from marine environments in association with other oxygenic phototrophs, which may have driven the niche-filling introduction of chlorophyll d. To investigate these unique adaptations, we have sequenced the complete genome of A. marina. The DNA content of A. marina is composed of 8.3 million base pairs, which is among the largest bacterial genomes sequenced thus far. This large array of genomic data is distributed into nine single-copy plasmids that code for >25% of the putative ORFs. Heavy duplication of genes related to DNA repair and recombination (primarily recA) and transposable elements could account for genetic mobility and genome expansion. We discuss points of interest for the biosynthesis of the unusual pigments chlorophyll d and α-carotene and genes responsible for previously studied phycobilin aggregates. Our analysis also reveals that A. marina carries a unique complement of genes for these phycobiliproteins in relation to those coding for antenna proteins related to those in Prochlorococcus species. The global replacement of major photosynthetic pigments appears to have incurred only minimal specializations in reaction center proteins to accommodate these alternate pigments. These features clearly show that the genus Acaryochloris is a fitting candidate for understanding genome expansion, gene acquisition, ecological adaptation, and photosystem modification in the cyanobacteria.     

Diverse high-torque bacterial flagellar motors assemble wider stator rings using a conserved protein scaffold

Although it is known that diverse bacterial flagellar motors produce different torques, the mechanism underlying torque variation is unknown. To understand this difference better, we combined genetic analyses with electron cryo-tomography subtomogram averaging to determine in situ structures of flagellar motors that produce different torques, from Campylobacter and Vibrio species. For the first time, to our knowledge, our results unambiguously locate the torque-generating stator complexes and show that diverse high-torque motors use variants of an ancestrally related family of structures to scaffold incorporation of additional stator complexes at wider radii from the axial driveshaft than in the model enteric motor. We identify the protein components of these additional scaffold structures and elucidate their sequential assembly, demonstrating that they are required for stator-complex incorporation. These proteins are widespread, suggesting that different bacteria have tailored torques to specific environments by scaffolding alternative stator placement and number. Our results quantitatively account for different motor torques, complete the assignment of the locations of the major flagellar components, and provide crucial constraints for understanding mechanisms of torque generation and the evolution of multiprotein complexes.Flagellated bacteria have tailored their motility to diverse habitats. For example, the enteric model organisms Salmonella enterica serovar Typhimurium and Escherichia coli colonize animal digestive tracts and can reside outside a host, assembling flagella over their cell body to swim. However, a diverse spectrum of flagellar swimming ability is seen across the bacterial kingdom. Caulobacter crescentus inhabits low-nutrient freshwater environments where it swims using a high-efficiency flagellar motor (1, 2), whereas Vibrio species produce high-speed, sodium-driven polar flagella to capitalize on the high sodium gradient of their marine habitat (3). On the other hand, the ε-proteobacteria and spirochetes, many of which thrive exclusively in association with a host, have evolved characteristically rapid and powerful swimming capabilities that enable them to bore through mucous layers coating epithelial cells or between tissues. Indeed, the ε-proteobacteria Campylobacter jejuni and Helicobacter pylori are capable of continued swimming in high-viscosity media that immobilize E. coli or Vibrio cells (4–6), and similar behavior is observed for spirochetes (7, 8).Despite differences in the organisms’ swimming ability, the flagellar motor is composed of a conserved core of ∼20 structural proteins (9). The mechanism of flagellar motility is conserved (10), with torque generated by rotor and stator components (9). Stator complexes, heterooligomers of four motility A (MotA) and two motility B (MotB) proteins, are thought to form a ring that surrounds the axial driveshaft. Transmembrane helices of MotA and MotB form an ion channel, and MotB features a large periplasmic domain that binds peptidoglycan (11, 12) and the flagellar structural component, the P-ring (13). The stator complex couples ion flux to exertion of force on the cytoplasmic rotor ring (the C-ring), which transmits torque to the axial driveshaft (the rod), universal joint (the hook), and helical propeller (the filament), culminating in propulsion of the bacterium. Biophysical (14) and freeze-fracture (15) studies together with modeling (16) have proposed that a tight ring of ∼11 stator complexes dynamically assembles around the rod above the outer lobe of the C-ring in closely related Salmonella and E. coli motors (which we collectively refer to as the “enteric motor”). However, despite these conclusions, and although the structures observed in subtomogram averages have been proposed to be the stator complexes (17–19), the locations and stoichiometries of the stator complexes remain to be confirmed.How can we explain the wide diversity in flagellar swimming abilities in the context of a conserved core flagellar motor? Biophysical studies suggest that the source of the difference lies, at least in part, in variations in the mechanical output of the motors themselves. Torques of motors from different bacteria have been shown to range over an order of magnitude, and torque correlates with swimming speed and the ability of bacteria to propel themselves through different viscosities, indicating that adaptations are likely to be at the level of the motor itself. [Torque also varies within a single species, up to a maximum value, as a function of the number of stator complexes incorporated into the motor (14)]. For example, C. crescentus motors have been measured to produce torques of 350 pN⋅nm (2). Estimates for the torque of the enteric motor ranges from ∼1,300 to ∼2,000 pN⋅nm (20, 21). The ε-proteobacterium H. pylori has been estimated to swim with torque of 3,600 pN⋅nm (22), and spirochetes are capable of swimming with 4,000 pN⋅nm of torque (21, 23). Sodium-driven motor torques in Vibrio spp. have been measured between ∼2,000 and 4,000 pN⋅nm (24), depending on the magnitude of the sodium gradient. It is noteworthy, however, that an estimated sodium motive force in Vibrio spp. that is lower than the standard E. coli proton motive force nevertheless drives the Vibrio motor with higher torque than the E. coli motor (24, 25), further suggesting that torque differences likely exist at the level of the motor. However, the molecular mechanism by which different motors might produce different torques has not been investigated.The simplest scenario for tuning motor torque would be evolved adaptation of motor architecture. In support of this scenario, we recently showed that many motors have evolved additional structures not found in the well-studied enteric motors (18), and we observed that the C-ring radius varies among species (17, 18). One of the most widespread novel structures is a periplasmic basal disk directly beneath the outer membrane, often co-occurring with varied uncharacterized additional structures, which we collectively term “disk complexes.” Consistently, disk complexes have been seen only in motors that produce torque higher than that in E. coli or Salmonella. For example, the sodium-driven ∼2,000+ pN⋅nm torque motors of Vibrio species assemble a disk complex featuring a basal disk beneath the outer membrane (18) in addition to smaller H- and T-rings composed of FlgOT (flagella O, T) and MotXY (motility X, Y), respectively (26, 27). It has been shown that the T-ring interacts with stator complexes in Vibrio spp. (28), although the exact location and number of stator complexes in Vibrio spp. remains unclear. ε-Proteobacteria such as Helicobacter species, C. jejuni, and Wolinella succinogenes also assemble disk complexes composed of large basal disks beneath the outer membrane together with additional smaller disks (18, 29). Although these and other cases of additional disks have been reported (18, 30), their relation to flagellar function remains enigmatic, and it is unclear if these widespread disk complexes are homologous or analogous.In this study, we hypothesized that bacteria have tuned their swimming abilities by evolving structural adaptations to their flagellar motors that would result in altered torque generation. Using electron cryo-tomography and subtomogram averaging, we found that Vibrio polar γ-proteobacterial and Campylobacter ε-proteobacterial flagellar motors incorporate 13 and 17 stator complexes, respectively, compared with the ∼11 in enteric bacteria. In both cases, these stator complexes are scaffolded into wider stator rings relative to the enteric motor by components of their respective disk complexes. The wider C. jejuni stator ring is further reflected in a considerably wider rotor C-ring. Further analysis of the components of the Vibrio and C. jejuni disk complexes reveals that they share a core protein, FlgP, but each has acquired diverse additional components to form divergent disk-complex architectures. We conclude by showing that our structural data of wider stator rings featuring additional stator complexes can quantitatively account for the differences in torque between different flagellar motors.     

The role of E255K/V-inclusive mutations in a Philadelphia-positive acute lymphoblastic leukemia with mutation evolution during sequential TKIs therapies: A case report

Rationale:Until recently, the survival rate in patients with Philadelphia-positive acute lymphoblastic leukemia (Ph+ ALL) was approximately 30%. Tyrosine kinase inhibitors (TKIs), which are a new class of drugs that target BCR-ABL fusion protein, have shown to be effective in treating Ph+ ALL in adults. However, the resistance mechanisms that promote the disease recurrence have altered the initial success of these revolutionary agents.Patient concerns:A 71-year-old Chinese female patient who suffered from severe shoulder and back pain for 1 week.Diagnosis:The patient was diagnosed with Ph+ ALL (B–cell) because of the following items. Complete blood count showed extremely abnormal white blood cell count (26.26×10⁹/l), hemoglobin concentration (65 g/l) and platelet count (14×10⁹/l). And because that Bone marrow aspirate showed 72.5% lymphoblasts and 59.30% lymphoblasts were confirmed by flow cytometry (FCM). At mean time, Real-time fluorescent quantitative PCR analysis confirmed that the P190 BCR/ABL fusion gene expression was 5.9%. Karyotype analysis indicated the following: 45, XX, −7, t (922) (q34; q11) [cp3].Interventions:The patient was treated with chemotherapy and different TKIs including imatinib, dasatinib, ponatinib, and bosutinib.Outcomes:The patient achieved complete remissions with different TKIs after diagnose but relapsed afterward and died of infection.Lessons:Multidrug-resistant mutations within the BCR-ABL1 kinase domain are an emerging clinical problem for patients receiving sequential TKIs therapy. Acquisition of E255K/V-inclusive mutations is usually associated with ponatinib resistance, thus it is necessary to screen out new real pan-inhibitor compounds for all BCR/ABL mutations and figure out the potential efficacy of asciminib-based drug combinations in the future.     

Surface plasmon mediates the visible light–responsive lithium–oxygen battery with Au nanoparticles on defective carbon nitride

Aprotic lithium-oxygen (Li-O₂) batteries have gained extensive interest in the past decade, but are plagued by slow reaction kinetics and induced large-voltage hysteresis. Herein, we use a plasmonic heterojunction of Au nanoparticle (NP)–decorated C₃N₄ with nitrogen vacancies (Au/N_V-C₃N₄) as a bifunctional catalyst to promote oxygen cathode reactions of the visible light–responsive Li-O₂ battery. The nitrogen vacancies on N_V-C₃N₄ can adsorb and activate O₂ molecules, which are subsequently converted to Li₂O₂ as the discharge product by photogenerated hot electrons from plasmonic Au NPs. While charging, the holes on Au NPs drive the reverse decomposition of Li₂O₂ with a reduced applied voltage. The discharge voltage of the Li-O₂ battery with Au/N_V-C₃N₄ is significantly raised to 3.16 V under illumination, exceeding its equilibrium voltage, and the decreased charge voltage of 3.26 V has good rate capability and cycle stability. This is ascribed to the plasmonic hot electrons on Au NPs pumped from the conduction bands of N_V-C₃N₄ and the prolonged carrier life span of Au/N_V-C₃N₄. This work highlights the vital role of plasmonic enhancement and sheds light on the design of semiconductors for visible light–mediated Li-O₂ batteries and beyond.

The aprotic lithium-oxygen (Li-O₂) battery promises ultrahigh theoretical energy density (∼3,600 Wh·kg⁻¹) and is operated with oxygen reduction to generate the product of Li₂O₂ and its reverse oxidation (2Li⁺ + O₂ + 2e⁻ ↔ Li₂O₂, E⁰ = 2.96 V) (1–5). The sluggish oxygen cathode reactions, including the oxygen evolution reaction (OER) and the oxygen reduction reaction (ORR), lead to a high discharge/charge overvoltage (∼1.0 V) during cycles and low round-trip efficiency (6–9). Since the pioneering work on the photoinvolved Li-O₂ battery using TiO₂ (10) or C₃N₄ (11) under ultraviolet (UV)-light irradiation, reduction of the charge/discharge overvoltage via a photomediated strategy has been extensively studied and is anticipated to solve the kinetic issues of the Li-O₂ battery (12–18). However, the light absorption of most semiconductors used is confined in the region of UV light, accounting for only ca. 4% of the solar spectrum (14–16). Expanding the light harvesting from UV to visible light is the long-term goal and challenge of photocatalysis (17–20). Simultaneously, high carrier recombination consumes the majority of photoelectrons and holes before catalyzing the targeted reactions, resulting in a mismatch between the carrier lifetime and kinetics of ORR or OER (19–21). This necessitates a structural design of semiconducting materials for visible-light harvesting to accelerate the cathode reactions in Li-O₂ batteries.Localized surface plasmon resonance (LSPR), which refers to the collective oscillation of conduction band (CB) electrons in metal nanocrystals under resonant excitation, has recently gained much attention (22–25). The decay of excited LSPR can produce hot electrons and holes, which initiate various chemical reactions (22, 23). Intriguingly, when plasmonic metal (e.g., Au, Ag) nanoparticles (NPs) come into contact with a semiconductor such as MoS₂, TiO₂, etc., an interfacial Schottky barrier forms; this barrier functions as a filter to force the energetic electrons or holes to migrate across the interface while inhibiting their reverse movement, thereby leading to effective electron–hole separation and suppressed charge–carrier recombination (26–30). LSPR systems generally are composed of plasmonic metal and semiconductors and exhibit the benefits of a low electron–hole recombination rate, enhanced light harvesting, and tailored response wavelengths from the visible to the near-infrared region (22). Recently, Au/CdSe (31) and Au/Ni(OH)₂ (32) heterojunctions have been attempted for a photocatalytic hydrogen evolution reaction and OER with the aid of hot electrons and holes under visible light. Coupling the plasmonic metal with suitable semiconductors for broadened light harvesting and a plasmon-enhanced effect is highly desirable for both ORR and OER in the Li-O₂ battery.Herein, we report defective C₃N₄ (Au/N_V-C₃N₄) decorated with plasmonic Au NPs as a bifunctional heterojunction catalyst that promotes cathode reactions of the Li-O₂ battery under visible light. The N_V on N_V-C₃N₄ is prone to adsorb and activate O₂, and the plasmon-excited electrons on Au migrate to the CB of N_V-C₃N₄ and relax to the N_V-induced defect band (DB) for O₂ reduction to LiO₂; then it undergoes electron reduction to Li₂O₂. Reversely, the Li₂O₂ is removed by the holes on the Au NPs driven by the applied voltage. The discharge voltage is raised to 3.16 V, and the charge voltage is lowered to 3.26 V at 0.05 mA·cm⁻² with a good rate capability and cycle stability. This investigation integrates a plasmonic heterojunction into the aprotic Li-O₂ battery and illustrates photoenergy conversion and storage under visible light.     

An ecocultural model predicts Neanderthal extinction through competition with modern humans

Archaeologists argue that the replacement of Neanderthals by modern humans was driven by interspecific competition due to a difference in culture level. To assess the cogency of this argument, we construct and analyze an interspecific cultural competition model based on the Lotka−Volterra model, which is widely used in ecology, but which incorporates the culture level of a species as a variable interacting with population size. We investigate the conditions under which a difference in culture level between cognitively equivalent species, or alternatively a difference in underlying learning ability, may produce competitive exclusion of a comparatively (although not absolutely) large local Neanderthal population by an initially smaller modern human population. We find, in particular, that this competitive exclusion is more likely to occur when population growth occurs on a shorter timescale than cultural change, or when the competition coefficients of the Lotka−Volterra model depend on the difference in the culture levels of the interacting species.Neanderthals are a human species (or subspecies) that went extinct, after making a small contribution to the modern human genome (1, 2). Hypotheses for the Neanderthal extinction and their replacement by modern humans, in particular as recorded in Europe, can be classified into those emphasizing competition with modern humans and those arguing that interspecific competition was of minor relevance. Among the latter are the climate change (3) and epidemic/endemic (4) hypotheses. However, an ecocultural niche modeling study has shown that Neanderthals and modern humans exploited similar niches in Europe (5), which, together with a recent reassessment of European Paleolithic chronology showing significant spatiotemporal overlap of the two species (6), suggests a major role for interspecific competition in the demise of the Neanderthals.Replacement of one species (or population) by another is ultimately a matter of numbers. One competing species survives while the other is reduced to, or approaches, zero in size. In the classical Lotka−Volterra model of interspecific competition, this process is called competitive exclusion (7). If Neanderthals were indeed outcompeted by modern humans, the question arises: Wherein lay the advantage to the latter species? Many suggestions have been made, including better tools (8), better clothing (9, 10), and better economic organization (11). These hypotheses share the premise that modern humans were culturally more advanced than the coeval Neanderthals.The purpose of our paper is threefold. First, we extend the Lotka−Volterra-type model of interspecific competition by incorporating the “culture level” of a species as a variable that interacts with population size (12, 13). Here, culture level may be interpreted as the number of cultural traits, toolkit size, toolkit sophistication, etc. Although, as noted above, many anthropological and archaeological discussions invoke interspecific cultural competition, there is, to the best of our knowledge, no mathematical theory of this ecocultural process. A mechanistic resource competition model is difficult to justify at present, because there is a limited understanding of “what the species are competing for… [or] how they compete” (14). Second, we use our interspecific cultural competition model to explore, analytically and numerically, the possibility that a difference in culture level, or in underlying learning ability, may produce competitive exclusion of a comparatively (although not absolutely) large regional (Neanderthal) population by an initially smaller (modern human) one. Third, we assume the competition coefficients of the Lotka−Volterra model to depend explicitly on the difference in the culture levels of the interacting species (rather than to be constants) and ask how this modification affects the invasion and subsequent dynamics.Dependence of the culture/technology level of a human population on its size has been the focus of many theoretical (15–21) as well as psychological (22–24), archaeological (25, 26), and ethnological (27–30) studies. However, the coupled dynamics of population size and culture level, where both quantities are treated as variables, has received less theoretical attention (12, 13, 31, 32).Taking refs. 12 and 13 as the point of departure, we extend previous treatments by introducing two such populations in direct competition with each other in the Lotka−Volterra framework. The two populations are described in terms of their size, N_i (≥0), their culture level z_i (≥0), i (=1, 2), and parameters to be defined below. We ask whether a population can be replaced by an initially smaller one, which has an advantage in culture level or in learning ability. This ecological perspective on the competition between “size−culture profiles” may inform ongoing debate on the replacement of Neanderthals by modern humans.     

Scaling up real networks by geometric branching growth

Real networks often grow through the sequential addition of new nodes that connect to older ones in the graph. However, many real systems evolve through the branching of fundamental units, whether those be scientific fields, countries, or species. Here, we provide empirical evidence for self-similar growth of network structure in the evolution of real systems—the journal-citation network and the world trade web—and present the geometric branching growth model, which predicts this evolution and explains the symmetries observed. The model produces multiscale unfolding of a network in a sequence of scaled-up replicas preserving network features, including clustering and community structure, at all scales. Practical applications in real instances include the tuning of network size for best response to external influence and finite-size scaling to assess critical behavior under random link failures.

In the context of network science, growth is most often modeled through the sequential addition of new nodes that connect to older ones in a graph by different attachment mechanisms (1, 2), including models of hidden variables, where nodes are characterized by intrinsic properties (3, 4). Other growth processes have also been considered, such as duplication to explain large-scale proteome evolution (5, 6). Here, we take an alternative approach and explore the relation between branching growth (7) and geometric renormalization (GR) (8) to explain self-similar network evolution. Renormalization in networks, based on the ideas of the renormalization group in statistical physics (9–11), acts as a sort of inverse branching process by coarse-graining nodes and rescaling interactions. Thus, branching growth can be seen as an inverse renormalization transformation: an idea that was introduced in ref. 12 using a purely topological approach to reproduce the structure of fractal networks, where fractality was interpreted as an evolutionary drive toward robustness. However, topological distances in networks are seriously constrained by the small-world property, while the characterization of fractality in real networks disregards fundamental features of their structure, including clustering and community organization.GR (8) is an alternative technique that can be performed by virtue of the discovery that the structure of real networks is underlain by a latent hyperbolic geometry (13, 14). Thus, the likelihood of interactions between nodes depends on their distances in the underlying space, via a universal connectivity law that operates at all scales and simultaneously encodes short- and long-range connections. This approach has been able to explain many features of the structure of real networks, including the small-world property, scale-free degree distributions, and clustering, as well as fundamental mechanisms such as preferential attachment in growing networks (4) and the emergence of communities (15, 16). Given a network map, GR produces a multiscale unfolding of the network in scaled-down replicas over progressively longer length scales. This transformation has revealed self-similarity to be a ubiquitous symmetry in real networks, whose structural properties remain scale-invariant as the observational resolution is decreased (8). This poses the question of whether this self-similarity could be related to the mechanisms driving the growth of real networks and, therefore, whether their evolution could be conceptualized within the framework of the GR group.In this work, we show that real networks—citations between scientific journals (17, 18) and international trade (19)—have evolved in a self-similar way over time spans of more than 100 y, meaning that their local, mesoscale, and global topological properties remain in a steady state as time goes by, with a moderate increase of the average degree. We demonstrate that the observations can be modeled by a geometric branching growth (GBG) process that produces a self-similar metric expansion. Beyond the capacity of the model to explain and predict the self-similar evolution of real networks effectively, the technique is flexible and allows us to generate scaled-up network replicas that, when combined with scaled-down network replicas (8), provide a full up-and-down self-similar multiscale unfolding of complex networks that covers both large and small scales. We illustrate the use of GBG multiscale unfolding in real network instances via the tuning of network size for optimal response to an external influence, referred to here as “the optimal mass,” and a finite-size scaling analysis of critical behavior under random link failures.