Hostname: page-component-cd9895bd7-gvvz8 Total loading time: 0 Render date: 2024-12-27T11:13:10.233Z Has data issue: false hasContentIssue false

Curbing the fruitfulness of self-replicating machines

Published online by Cambridge University Press:  08 July 2022

Alex Ellery*
Affiliation:
Department of Mechanical & Aerospace Engineering, Carleton University, 1125 Colonel By Drive, Ottawa, ON. K1S 5B6, Canada
*
Author for correspondence: Alex Ellery, E-mail: [email protected]
Rights & Permissions [Opens in a new window]

Abstract

The self-replicating machine has high utility by virtue of its universal construction properties and its productive capacity for exponential growth. Their capacity is unrivalled. They can be deployed to the Moon to industrialize it using local in-situ resources in the short term to open up the solar system and thence deployed on interstellar spacecraft to explore the entire Galaxy by exploiting in-situ stellar system resources. Nevertheless, there are significant concerns regarding the inherent safety of self-replicating machines. We consider the general problem of runaway population growth in physical self-replicating machines to prevent the grey goo problem, the number of offspring spawned by self-replicating machines may be controlled at a genetic level. We adopt a biologically-inspired approach based on telomeres, DNA endcaps that are progressively shortened during cellular replication. This acts as a counter that imposes a limit to the number of replication cycles (Hayflick limit). By examining the biological process in detail, we can obtain some insights in implementing similar mechanisms in self-replicating machines. In particular, we find that counting mechanisms are vulnerable to cancerous runaway.

Type
Research Article
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
Copyright © The Author(s), 2022. Published by Cambridge University Press

Introduction

We have been developing aspects of a physical self-replicating machine for deployment on the Moon using only local in-situ resources (Ellery, Reference Ellery2015a, Reference Ellery2017a). Self-replicating machines deployed on the Moon exploit their exponential growth capability to rapidly construct a production-based industrial infrastructure to support the low-cost settlement of the Moon (Ellery, Reference Ellery2016). We have examined the lunar resource requirements and the required processing of lunar raw materials within a lunar industrial ecology in detail (Ellery, Reference Ellery2020; Ellery et al., Reference Ellery, Mellor, Wanjara and Conti2022). The self-replicating machine we consider here has grander applications – interstellar spacecraft with a self-replicating payload can explore the entire Galaxy at a low-cost (Freitas, Reference Freitas1980). Indeed, it might be considered a natural extension from self-replicating machines within the solar system to beyond with minimal cost. We have been 3D printing crucial components to demonstrate universal construction properties. We have demonstrated 3D printing of electric motors which are a key component of kinematic machines (Elaskri and Ellery, Reference Elaskri and Ellery2020). We have yet to 3D print active computational devices but this activity is ongoing (Prasad and Ellery, Reference Prasad and Ellery2020). We have explored the possibility of a bio-inspired control architecture for the industrial ecology based on genetic regulatory networks (Ellery, Reference Ellery2021). The self-replicating machine has high utility by virtue of its universal construction properties and its productive capacity for exponential growth which suggest inevitability. However, there are significant concerns regarding the inherent safety of self-replicating machines primarily associated with its exponential growth potential – this is the grey goo scenario. We have explored several approaches to this problem (Ellery and Eiben Reference Ellery and Eiben2019) including (i) the implementation of kill switches (through the salt and tunicose contingencies) that deny centralized supplies of resources; (ii) the implementation of error detection and correction codes (EDAC) at a genetic (memory) level to prevent evolutionary divergence. Only the latter would be applicable to a self-replicating interstellar probe. But there are further, more subtle biomimetic approaches that we explore here relating to maturation such as ageing and cancer.

Problem of uncontrolled self-replication

Humankind blessed them, and Humankind said to them, ‘Be fruitful and multiply; but fill not the earth nor subdue it; you have no dominion over the fish of the sea, nor over the birds of the air, nor over any living thing that moves on the earth or elsewhere.’ Then Humankind said, ‘Behold, we have supplied you the element-yielding minerals of the Moon, and the rocks in which are the element-yielding minerals; to you it is food…. but consume frugally’. Forgive the literary adulteration but it forcibly illustrates the immense power and implicit responsibility bestowed upon humanity in pursuing self-replicating machine technology. One of the thorniest problems in developing self-replicating machines is the fear of an unrestrained population explosion – the grey goo scenario. It is imperative that this problem is addressed in a robust manner that satisfies not only the real threat but also the perceived threat (Kahneman and Tversky, Reference Kahneman and Tversky1979) which, frankly, is magnified out of all proportion (Joy, Reference Joy2000) but nevertheless must be allayed. It has been suggested that runaway self-replication of machines on Earth will yield at least a 4 °C global temperature rise within 2 years which would be readily detectable (Freitas, Reference Freitas2001). Similarly, the release of self-replicating machines throughout the Galaxy has been deemed as irresponsible that no intelligent species would undertake (Sagan and Newman, Reference Sagan and Newman1983). It is worth noting that uncontrolled replication is not inherently undesirable – it is the end to which it is deployed that is the determinant of its ethical character. Uncontrolled self-replication has been proposed to spread vaccines through self-spreading viruses as a vector (Lentzos et al., Reference Lentzos, Rybicki, Engelhard, Paterson, Sandholtz and Reeves2022). In this case, viral evolution is suppressed by limiting viral lifetimes. However, the efficacy of such genetic manipulation is unproven in the wild – a dynamic, complex and uncontrolled environment. Nevertheless, the sheer utility of self-replicating machines renders it unlikely that self-denial of such technology is realistic – its universal construction and exponential growth capabilities render it extremely high economic utility (Ellery, Reference Ellery2017b). It behoves us therefore to address the risk of proliferation directly.

In previous work, we have addressed some issues regarding the control of self-replicating machine populations (Ellery and Eiben, Reference Ellery and Eiben2019) – we had specifically addressed the implementation of genetic redundancy and error correction codes to prevent evolutionary variance as the population grows through the self-replication process. One safeguard that might be implemented is restrictions on offspring population. In this paper, we continue to address the problem of control of self-replicating machine populations by focussing on imposing limits on the number of offspring that any individual self-replicating machine can spawn – a Hayflick limit – and inherent (rather than external) kill switch approaches. We shall examine in detail how biological constraints on cellular division are imposed to explore if there are biomimetic lessons that we can learn in implementing our technological solutions. Our focus will be on the implementation of counters and their biological counterpart, the telomeres.

How biological cells count

Prokaryotes demonstrate that biological organisms can propagate indefinitely with continuous repair without suffering senescence. Unlike prokaryotes in which DNA is circular however, eukaryotic DNA is packaged into linear chromosomes with free ends requiring telomere endcaps. The relationship between Hayflick limits and the concept of telomeres, telomerase and senescence was first described by Alexei Olovnikov (Reference Olovnikov1996). Telomeres are protein-DNA complexes at the ends of eukaryotic chromosomes that act as buffers by forming molecular caps to protect them. They are characterized by non-coding sequences of many short tandem repeats of species-specific motifs. These motifs comprise 5–8 base pairs including G-rich sequences at the 3’ end of the chromosome, e.g. the 6 base-pair (TTAGGG)n in vertebrates (Neidle and Parkinson, Reference Neidle and Parkinson2003). They are usually identical tandem repeats but there can be variants. Telomeric repeats are highly conserved in different eukaryotic groups from yeast to humans (Moyzis et al., Reference Moyzis, Buckingham, Cram, Dani, Deaven, Jones, Meyne, Ratliff and Wu1988). The average length of telomeric repeats varies between species but the average length correlates with variability rather than genome size – in humans, telomere length is 2–15 kbases but in mice, it is >30 kbases.

Most of the G-rich strand is duplex but a single (lagging) strand protrudes 100–300 bases (in humans) beyond the complementary C-rich (leading) strand, i.e. the G-rich strand is longer than the complementary C-rich strand resulting in a G-rich 3’ overhang (G-tail) of 100–300 single-stranded repeats which binds to part of the ds region forming a (telomere) t-loop. The telomeric repeats are primarily double stranded (TTAGGG/CCCTAA)n and the single-stranded overhang (TTAGGG)n is much shorter than the double strand. It is the single-stranded DNA extension that differentiates the telomere endcaps from double-strand breaks (DSB). Double strand breaks during replication are detected and processed through S-phase checkpointing. Single strands prevent telomeres from being treated as DSB, especially nonhomologous end joining (EJ) of the broken ends (McEachern et al., Reference McEachern, Krauskopf and Blackburn2000). Non-homologous EJ involves the DNA ligase and the DNA-PK kinase protein binding to the DNA end. The single-stranded G-tail overhang forms a terminal (t) loop structure at the end of the double-stranded telomeric repeats which acts to protect the telomeres (Griffith et al., Reference Griffith, Comeau, Rosenfeld, Stansei, Bianchi, Moss and de Lange1999). The tandem repeats serve as binding sites for specific proteins and repressor sites for nearby gene expression. Shelterin is a protective protein complex comprising six telomere-binding proteins that bind to telomeric repeats and prevent double-strand break repair mechanisms. In mammals, the six telomere-binding proteins are a core of TRF1 and TRF2 (telomere repeat binding factors) which attract RAP1 (TRF1-interacting protein), TIN2 (TRF1-interacting factor), POT1 (protection of telomeres 1) and TPP1 (adrenocortical dysplasia homologue). They bind to both double and single-strand regions of the telomere. TRF1, TRF2 and POT1 bind to double-strand telomeres with high specificity; POT1 binds to the single strand overhang; TN2 tethers TRF1 and TRF2; TPP1 tethers POT1 to TIN2. TRF1 inhibits telomere elongation but TRF2 renders telomeric ends into t loops thereby protecting against double-strand break response of senescence or apoptosis (Kierszenbaum, Reference Kierszenbaum2000). The t loop is formed by the 3’ single-stranded overhang bonding with the double-stranded telomeric repeats in the presence of TRF2 and requires sufficient double-strand length for t loop formation (Collins, Reference Collins2000). T loop formation requires TRF2 that renders the chromosome end non-sticky to p53-dependent DNA damage repair activity. In fact, there are a host of molecules involved in telomere function such as CST which comprises three proteins (Venkatesan et al., Reference Venkatesan, Khaw and Hande2017).

During telomere replication, the C strand replicates through lagging strand synthesis and the G strand through leading strand synthesis. DNA polymerase synthesizes the continuous DNA leading strand in the 5’-to-3’ direction. The lagging strand of the replication fork requires discontinuous Okazaki fragments to be joined by laying a short 10–16 nucleotide RNA primer to initiate DNA synthesis. Once the RNA primer is removed at the 3’ end, 50–200 bp of the lagging strand end cannot be replicated. 5’-to-3’ DNA synthesis by DNA polymerase leaves the overhanging G-rich strand of DNA (TTAGGG)n at the 5’ end incompletely replicated because the lagging C-rich strand is recessed. Leading strand DNA replication halts prematurely because of a lack of a template for the 3’ overhang resulting in incomplete replication at telomeric chromosome ends (Kelleher et al., Reference Kelleher, Teixerira, Forstemann and Lingner2002). The double-strand telomeric DNA constituting the bulk of telomeric DNA is packaged into chromatin and replicated; the single strand telomeric DNA is not packaged into chromatin and is not replicated. Hence, at each cell division, the single strand 3’ end of the telomere is not replicated causing telomere shortening of 50–200 bases per replication cycle. Thus, telomeres shorten progressively from their maximum length of 10–15 kb to a critical length of 4–6 kb in size, whence cell senescence is invoked. Cell senescence is invoked by DNA damage, telomere loss, de-repression of cyclin-dependent kinase inhibitor (CDKN2a) and other cellular stressors. Indeed, levels of CDK2Na expression are a good indication of cellular ageing. This limits the number of times cells can replicate to around 50 ± 10 cell divisions (varying in culture from 80–90 times in newborn tissue to 20–30 times in elderly tissue) – this is the Hayflick limit whence cellular senescence ensues (Hayflick, Reference Hayflick1984) though cellular senescence typically occurs long before this limit (of ~1015 cells after 50 doublings) is reached (Hayflick, Reference Hayflick1979). Accelerated ageing syndromes such as Werner's syndrome and Trisonomy 21 are characterized by reduced replicative capacity. Indeed, it has been demonstrated that telomere shortening acts as a clock that causally induces replicative senescence (Chiu and Harley, Reference Chiu and Harley1997; Bodnar et al., Reference Bodnar, Ouellette, Froklis, Holt, Chiu, Morin, Harley, Shay, Lichsteiner and Wright1998). Hence, telomere shortening acts as a mitotic replication copy counter. When a few telomeres become critically short, the G-rich binding sites for proteins are no longer functional in telomere replication and end capping. Short telomeres even occur in tissues with low replication rates. As well as replicative cycle losses of telomeres, there are also random telomere losses due to oxidative damage to sensitive G-rich DNA of telomeres (Lansdorp, Reference Lansdorp2005). Ageing correlates with diminished DNA repair capacity of regulatory telomeres which accumulate damage (Kruk et al., Reference Kruk, Rampino and Bohr1995).

There is a mechanism of alternative lengthening of telomeres (ALT) in which DNA is copied from telomere to telomere through homologous recombination (HR) using telomeres as copy templates (Dunham et al., Reference Dunham, Neumann, Fasching and Reddel2000). This is a type of recombination involving gene conversion that maintains extremely long telomeres (Kass-Eisler and Greider, Reference Kass-Eisler and Greider2000). However, we do not pursue this mechanism here.

In the germline cells (and cancers) but not healthy somatic cells, telomeres may be maintained indefinitely by telomerase. Germline cells are effectively immortal due to telomerase activation – telomeric DNA extension is coupled to DNA replication during the S phase. Telomerase is a specialized ribonucleoprotein (RNP) that binds to telomeric DNA. It incorporates a reverse transcriptase enzyme subunit but, unlike conventional reverse transcriptases which are pure proteins, includes an RNA subunit (Stone, Reference Stone2018). Telomerase is a RNP enzyme comprising the hTERT (human telomerase reverse transcriptase) protein and hTERC (human telomerase RNA component) RNA (Blackburn, Reference Blackburn1992). The hTERT enzyme catalyses the DNA sequence repeats from the TERC RNA template. TERT has highly conserved reverse transcriptase protein signatures because it generates DNA from its non-coding RNA template. Reverse transcriptase evolved from RNA-dependent RNA polymerases while telomerases originated with ancestral eukaryotes from retrotransposons (Nakamura and Cech, Reference Nakamura and Cech1998). TERC is the non-protein-encoding RNA transcribed by RNA polymerase II and is polyadenylated. The human telomerase RNA segment is 451 bases in length containing an 11-base RNA region 5’…CUAACCCUAAC…3’ (Simonsson, Reference Simonsson2003). It is this RNA sequence of telomerase that acts as a template from which the telomere DNA tandem repeats are copied. Telomerase adds new single-stranded telomeric tandem repeats to the 3’ chromosome end using its RNA template from which it only generates and elongates G-rich DNA strands. Complementary strand synthesis completes the double-strand segments up to the overhang. Telomerase requires a DNA primer onto which telomeric repeats are added in the usual 5’ to 3’ direction. The DNA polymerase-primase complex is crucial for telomere maintenance through elongation of multiple repeats. Telomerase thus lengthens chromosomes with repetitive sequences to compensate for telomere shortening. Regulation of transcription of hTERT is the limiting step in telomerase activation (Aisner et al., Reference Aisner, Wright and Shay2002). hTERT mRNA exhibits differential splicing into different tissue-specific splices that regulates the amount of telomerase enzyme.

Telomerase adds 390–420 base pairs of telomeric repeats to a telomere while oxidative stress can shorten telomeres by 100–600 base pairs per deletion depending on the degree of oxidative stress so cellular lifespan enhancement requires oxidative attrition to be below ~400 base pair deletions (Arkus, Reference Arkus2005). Mitochondrial DNA is essential for oxidative phosphorylation in the eukaryotic cell – cellular ageing is associated with mutations in mtDNA with high ageing rates in cells with high energy demands such as muscle and nerve cells (Nagley and Wei, Reference Nagley and Wei1998). However, there is no direct relationship between longevity in mammals and metabolic rate based on the assumption that ageing results from the accumulation of molecular damage due to genetic errors or toxic metabolites (Austad and Fischer, Reference Austad and Fischer1991). TERT regulates the level of mitochondrial oxygen radicals (Ait-Aissa et al., Reference Ait-Aissa, Ebben and Kadlec2016). DNA integrity checkpoints control the balance between telomeric shortening and lengthening (Chakhparonian and Wellinger, Reference Chakhparonian and Wellinger2003). Telomere homoeostasis implements a dynamic balance between lengthening and shortening which maintains a telomere length of 9–15 kbp in length (Bodnar et al., Reference Bodnar, Ouellette, Froklis, Holt, Chiu, Morin, Harley, Shay, Lichsteiner and Wright1998). Telomeres are switched either open or closed to telomerase activity. Cdc13p recognizes a minimal 11 nucleotide single strand sequence as a G-rich overhang (t-loop) and initiates telomerase activity by opening the telomere. Telomere maintenance involves several steps (Dubrana et al., Reference Dubrana, Perrod and Gasser2001): (i) telomere unfolds so that the tree end is open to enzymes; (ii) telomerase and DNA replication together synthesize the C-rich strand regulated by length-sensing feedback and second-strand synthesis; (iii) telomere refolds into its protective terminal.

Once a critical shortness of telomere length is reached in germ cells (but not somatic cells), telomerase activity can be activated during the late S phase of the cell cycle. It is the shortest telomeres rather than the average length that determine cellular response (Armanios and Blackburn, Reference Armanios and Blackburn2012). Telomerase is expressed only in germ cells but not somatic cells except during early embryonic development in somatic tissue (Wright et al., Reference Wright, Piatyszek, Rainey, Byrd and Shay1996). Embryonic stem cells possess very long telomeres that lengthen due to high TERT expression and adult stem cells also exhibit telomerase activity but this diminishes with adult stem cell age (Flores and Blasco, Reference Flores and Blasco2010). Indeed, during development, there is a major shift in telomere behaviour (Bekaert et al., Reference Bekaert, Derradj and Baatout2004). During early development of the germline, telomere length is maintained by balancing shortening induced during cell division against lengthening by telomerase which becomes particularly active after fertilization. During later differentiation into specialized somatic cells, prior to the differentiation of sex, shortening begins to dominate as telomerase activity is repressed, starting the molecular clock. Hence, telomere shortening contributes to ageing. As ageing occurs, telomerase activity declines. If telomerase is overexpressed, telomere slippage during cell replication causes excessive replication resulting in cancer.

When biological counting stops

There are mechanisms for genome maintenance to prevent mutation into oncogenes including multiple repair mechanisms all of which are highly conserved across evolutionary timescales (Hoeijmakers, Reference Hoeijmakers2001). There are four main repair systems in mammals – nucleotide excision repair (NER), base excision repair (BER), HR and EJ. NER repairs helix-distorting lesions that affect base pairing and prevent transcription. BER repairs chemical changes in bases. Many environmental insults yield characteristic genetic mutations, e.g. UV light converts the base cytosine into thymine and are characteristic of melanomas. The less frequent DSB which are more dangerous than single-strand breaks are repaired through HR which generates a second copy of DNA for aligning breaks and EJ when there is no second copy. NER is the most versatile of repair mechanisms with two subpathways – the global genome is scanned for helix-distorting damage and transcription-coupled repair focuses on the damage that blocks RNA polymerase elongation. HJ and ER are the main repair modes, however. Paediatric cancers are rare but concentrated in the brain, bone and white blood cells, all of which have undergone rapid recent evolutionary change or subject to strong host-parasite selection, suggesting that evolutionary disequilibria have not yet been corrected. Carcinogenesis is an evolutionary process that involves natural selection between genetic variations in somatic cancer cells, especially those that eliminate genetic repair mechanisms (Crespi and Summers, Reference Crespi and Summers2005). Such genomic instability with inactivation of DNA repair pathways is characteristic of the development of cancer malignancy (Tubbs and Nussenzweig, Reference Tubbs and Nussenzweig2017). The tissue organization theory regards cell proliferation as the natural state of all cells (Soto and Sonnenschein, Reference Soto and Sonnenschein2004). This places cellular replication as a fundamental default process in cellular life because it has evolved so – this will not be the case for a self-replicating machine.

If DNA repair cannot restore genetic integrity, there are two main cellular responses to damage – cellular senescence and programmed cell death (such as apoptosis), the latter killing the cell and the former irreversibly preventing cell replication. They are both invoked by common triggers such as DNA damage or oncogenic stress and involve common regulatory mechanisms such as the p53 pathway. p53 is activated by short telomeres initiating cellular senescence or apoptosis depending on the tissue. The default state of the p53 gene is inhibited by protein MDM2 to low-level activity until environmental stresses decouple MDM2 (Eisenstein, Reference Eisenstein2022). This activates p53 to regulate the expression of a cascade of gene products that invoke apoptosis. However, some tumours produce excess MDM2 even without p53 mutations. Cellular senescence is a stable cell cycle arrest that occurs due to cellular damage such as DNA mutation or telomere loss. Senescence is characterized by a distinct secretion profile of inflammatory cytokines and degraded proteins due to cellular damage which determines the response of neighbouring cells. Cell senescence induced by critically-shortened telomeres is the causative agent of ageing and its associated diseases (Xi et al., Reference Xi, Li, Ren, Zhang and Zhang2013). The characteristic feature of cellular senescence is the halting of cell division. It occurs when telomere uncapping invokes the loss of the single-strand overhang which activates double-strand break repair processes. Cellular senescence is initiated when telomeres still retain several kbases of tandem repeats due to p53 and p21 activation (Wright and Shay, Reference Wright and Shay1995). It is also a response to a wide variety of stresses including DNA damage, oxidative exposure, tumorigenesis, etc in which cells halt cell division for tumour suppression (Ben-Porath and Weinberg, Reference Ben-Porath and Weinberg2005). Cancer is also linked to cell senescence in which cell senescence blocks tumour formation (Finkel et al., Reference Finkel, Serrano and Biasco2007). Cellular senescence also occurs in embryonic development and is dependent on p21 regulated by the tumour necrosis factor pathway (Munoz-Espin et al., Reference Munoz-Espin, Canamero, Maraver, Gomez-Lopez, Contreras, Murillo-Cuesta, Rodriguez-Baeza, Varela-Nieto, Ruberte, Collado and Serrano2013).

Programmed cell death occurs in all multicellular animals, of which apoptosis is the most prominent. Apoptosis is an essential biochemical pathway for metazoans especially in embryonic development, adult tissue homoeostasis and defence against mutation (Danial and Korsmeyer, Reference Danial and Korsmeyer2004). Apoptosis is an altruistic cell behaviour that induces cell suicide (programmed cell death) without inflammation (a defining characteristic of necrotic cell death) in response to noxious stimuli. For example, in neurodegeneration, neuron apoptosis is induced by oxidative stress (Krantic et al., Reference Krantic, Mechawar, Reix and Quirion2006). Psychological stress is associated with higher oxidative stress which reduces telomerase activity and shortens telomerase length (Epel et al., Reference Epel, Blackburn, Lin, Dhabhar, Adler, Morrow and Cawthorn2004). Similarly, epithelial cells die through apoptosis when detached from other epithelial substrate cells – carcinomas result when apoptosis is suppressed. While apoptosis is a genetically-controlled, well-ordered process, necrosis is reactively invoked by the leakage of protein-digestive enzymes (cathepsins) from ruptured intracellular lysosomes due to increased intracellular Ca2+ ions.

There are two major signalling pathways for apoptosis – death receptor pathway and mitochondrial pathway (Gupta, Reference Gupta2001). The death receptor pathway includes the CD95 death pathway and the tumour necrosis factor pathways which are associated with immune response through T cell proliferation. Each T cell carries a distinct T cell receptor that recognizes a specific antigen presented at the surface of an invading cell. The thymus generates a wide variety of immature T cells (thymocytes) which undergo selection through programmed cell death. During apoptosis, the cell cytoplasm shrinks in volume followed by organelle and chromatin condensation while the cell membrane bulges. The DNA and lipids are packaged to enable neighbouring cells to recycle these resources through phagocytosis (Bar, Reference Bar1996). DNA is cleaved into fragments of 180–200 bp in length by Ca2+-dependent endonucleases during apoptosis rather than the random cleaving by lyzosomes that occurs during necrosis. In fact, this internucleosomal cleavage between histones ~180–200 bp in length is one type of cleavage – there are others including fragmentation into large 50–300 bp segments prior to internucleosomal cleavage corresponding to chromatin loop domains, and single-strand cleavage (Bortner et al., Reference Bortner, Oldenburg and Cidlowski1995). There are at least two endonucleases with different mechanisms involved in DNA fragmentation during apoptosis – caspase-dependent DFF (DNA fragmentation factor) and caspase-independent mitochondrial EndoG (endonuclease G) (Zhang and Xu Reference Zhang and Xu2002). Caspase 1, caspase 4 and caspase 5 promote lytic cell death triggering inflammation. Caspase 8 promotes cell death by inflammation. The destruction of DNA suppresses inflammatory response. The cell membrane pinches off to form multiple internal vesicles containing fragments of the cell contents. Apoptotic cells secrete lysophospatidylcholine (LPC) chemicals that invite phagocytes. Apparently, the situation is more complex as reduced apoptosis can prevent cancer progression by reducing cell division and so mutations (Luria-Delbruck theory) (Wodaerz and Komarova, Reference Wodaerz and Komarova2007).

When biological counting runs amok

Ageing is primarily caused by DNA damage (Schumacher et al., Reference Schumacher, Pothof, Vijg and Hoeijmakers2021) which induces (i) genome instability that prevents transcription and regulation despite constant repair mechanisms, (ii) telomere dysfunction due to progressive rather than critical shortening, (iii) epigenetic changes due to modifications of DNA methylation and polyADP-ribosylation of histones, (iv) proteostatic stress resulting in protein misfolding and (v) mitochondrial dysfunction due to reactive oxygen species damage from oxidative respiration for which there are fewer repair mechanisms. Such DNA damage invokes cell senescence which evolved as a mechanism to prevent cell over-proliferation (cancer). It is unlikely that the average human lifespan can extend beyond 85 years unless cancers can be controlled (Olshansky et al., Reference Olshansky, Carnes and Desesquelles2001), and even if this is the case, such longevity increases are likely to be modest (Turner, Reference Turner2004). Cancers are due to the accumulation of somatic mutations resulting in around 30% of deaths in developed countries. Cancer is ubiquitous among metazoans including mammals – although higher cancer rates occur in larger, long-lived organisms within species as expected, there no is evidence of such correlation between different species (Peto's paradox).

There are four main cancer markers: (i) evasion of apoptosis via mutation; (ii) expression of telomerase inducing limitless replication; (iii) angiogenesis that grows tumour blood supplies; (iv) tissue invasion and metastasis which causes 90% of cancer deaths. Cancer involves a succession of distinct genetic mutations (most commonly acquired rather than inherited) in specific tumour suppressor genes and proto-oncogenes that convert normal cells into cancer cells that proliferate into malignant tumours due to their resistance to cellular apoptosis (Fadeel and Orrenius, Reference Fadeel and Orrenius2005). Genetic mutations can affect interdependent signalling pathways that interact in multiple ways contributing to multiple steps in the development of cancer (McCormick, Reference McCormick1999), e.g. there are multiple cellular checkpoints in metabolic pathways such as Ras, p53, p16, etc which must be circumvented in tumourigenesis. Tumour suppressor genes slow cell division while proto-oncogenes aid cell growth. Oncogenes are cellular genes that have been captured into viral genomes through reverse transcriptase, e.g. src gene of Rous sarcoma virus that infects chickens. Proto-oncogenes are commonly involved in cell replication and differentiation. Oncogenes can inhibit autophagy which acts as a tumour suppressor. Autophagy is a coordinated series of processes that are fundamental to anti-ageing. Oncogene activation is a major driving force in the development of cancer. They generate stalled replication forks which tend to collapse into DSB. There are multiple processes through which normal proto-oncogenes are converted into active oncogenes (Land et al., Reference Land, Parada and Weinberg1983). If the mutation occurs in a tumour suppressive gene that controls cell division or other crucial apoptotic genes, then the cell does not undergo apoptosis and can allow cancer to emerge. Proto-oncogenes mutate in cancer-generating oncogenes that remain switched on. While oncogenes are activated (switched on) through chromosomal rearrangements or gene duplication, tumour suppressor genes such as TP53 are switched off (occurs in greater than 50% of human cancers).

Telomerase activity is repressed in somatic cells but is reactivated in cancer cells resulting in tumour growth (Kim et al., Reference Kim, Piatyszek, Prowse, Harley, West, Ho, Coviello, Wight, Weinrich and Shay1994). The majority of cancer cells exhibit high telomerase activity suggesting that senescence through telomere shortening is a malignant tumour-suppression mechanism that is activated in response to oxidative stress preventing telomerase expression (Vixtorelli and Passos, Reference Vixtorelli and Passos2017). Telomerase activity is present in 85–90% of all human tumours. Cancer is usually associated with shorter telomeres ~2–4 kbases than in healthy cells in conjunction with telomerase activation (Autexier and Greider, Reference Autexier and Greider1996). Tumour cells initially reproduce rapidly, generating shortened telomeres. Genomic instability arises due to the shortest telomeres rather than the average telomere length increasing the prospect of tumour growth (Lundblad, Reference Lundblad2001). Such short telomeres indicate DNA damage and initiate cancer. Only subsequently, short telomeres are subsequently maintained through telomerase activation. Hence, telomeres in cancer cells are smaller than in normal cells because cancer cells synthesize telomerase only after the onset of uncontrollable replication. Cancer cells become immortalized through the activation of hTERT telomerase expression to actively maintain stable telomere lengths (Masutomi et al., Reference Masutomi, Yu, Khurts, Ben-Porath, Currier, Metz, Brooks, Kaneko, Murakami, DeCaprio, Weinberg, Stewart and Hahn2003). Telomerase (TERT) suppression can lead to tumour cell apoptosis or senescence. Telomerase inhibitors such as PNA, hammerhead ribozymes or antigens could potentially be deployed against telomerase-mediated cancers (White et al., Reference White, Wright and Shay2001). Oncogenes such as Myc and Wnt act as transcriptional regulators of telomerase. Telomerase then allows further replication but with shortened telomeres. Cancers are caused by mutations in DNA repair genes that cause deficient DNA repair enzymes and so accelerate mutation rates. Telomeres on the short arm of chromosome 17 are shorter than telomeres on other human chromosomes on average – this chromosome includes the p53 gene.

Tumour suppression genes regulate a range of cell processes including p53 transcription factor which regulates cell cycle checkpoints as the genomic gatekeeper (Sherr, Reference Sherr2004). p53 offers two broad functions – it induces cell senescence responses to stress and regulates metabolic activity. Cell senescence is controlled by tumour-suppressor pathways of which the most prominent in human cancers are p53 and pRB which also interact with each other in several ways (Campisi, Reference Campisi2001). Tumour suppressor genes prevent abnormal cell growth by inducing repair, senescence or programmed cell death. The TRP53 gene which encodes the p53 protein is a tumour suppressor and exhibits the commonest mutation in human cancers. Because p53 also regulates glucose metabolism, this prevents cancer from upregulating glycolysis and so glucose consumption required for cell proliferation is mediated by activating AMP-activated protein kinase (Warburg effect) (Bensaad and Vousden, Reference Bensaad and Vousden2007). It regulates the transcription of TIGAR (TP53-induced glycolysis and apoptosis regulator) and the SCO2 (synthesis of cytochrome c oxidase 2) proteins involved in glucose and mitochondrial oxidative respiration respectively. p53 activity resists cancer but promotes ageing. pRB regulates transcription more indirectly through interaction with transcription factors. Activation of p53 transcription factor positively regulates transcription of p21, a cyclin-dependent kinase inhibitor that causes cell cycle arrest characteristic of cell senescence (Sahin and DePinho, Reference Sahin and DePinho2010). Both p53 and p16 must be inhibited for cell proliferation to occur.

Somatic mutations in the p53 gene are frequent (up to 50%) in most cancers and germline mutations in the p53 gene causes Li-Fraumeni syndrome which predisposes to a large variety of early cancers – most such mutations are multiple incidences of single amino acid changes rather than frameshift or nonsense mutations, 25% of which are C:G > T:A substitutions at CpG sites (Olivier et al., Reference Olivier, Hollstein and Hainaut2010). p53 is a tumour suppressor gene activated by DNA damage that induces arrest of the cell cycle in the G1 phase for DNA repair processes and/or apoptosis. If the p53 gene is mutated, cancer cells are permitted to survive and proliferate without restraint. p53 upregulates Bid, Bax and Bad, three of several proteins of the bcl-2 family that activate apoptosis by decreasing the stability of phospholipid bilayers of the outer mitochondrial membrane by forming channels through it. Genetic instability arises from several sources but most notably from deactivated DNA mismatch repair gene (MSH2 and MLH1) mutations generating mutation rates two or three orders of magnitude higher than in normal cells (Cahill et al., Reference Cahill, Kinzler, Vogel stein and Lengauer1999). If genes for DNA repair enzymes mutate, mutational errors accumulate at a much higher rate leading to cancers (Pray, Reference Pray2008). Germline mutations in mismatch repair genes are associated with cancer predisposition in humans including colon cancers (Radman et al., Reference Radman, Matic, Halliday and Taddei1995; Peltomaki, Reference Peltomaki1997). As replication proceeds uncontrollably in cancers, tumours grow and invade neighbouring tissue, sometimes detaching and travelling to other distant tissues and generating new tumours. This is metastasis.

Some 10–15% of cancers are not characterized by telomerase activity but exhibit extremely long telomeres maintained through ALT (Bryan et al., Reference Bryan, Englezou, Dalla-Pozza, Dunham and Reddel1997; Blasco and Hahn, Reference Blasco and Hahn2003). ALT occurs in most cases of solid tumours (10–15% of cancers) which recombine telomeres from different chromosomes. ALT is believed to have derived from homologous repair. ALT-based tumours cannot be treated with telomerase inhibitors. ALT cancers have poor prognosis and telomerase-based cancers can switch to ALT cancers under selection due to therapeutic intervention.

There are epigenetic effects in cancer growth that do not involve underlying genetic mutations. Epigenetic phenomena include methylation and histone modification. DNA methylation involves reversible attachment of a methyl group to a cytosine nucleotide located next to a guanine nucleotide (CpG site). Methylated CpG sites act as binding sites for proteins that alter DNA's structure which prevents transcription (with CpG/CpC ratio>0.6), i.e. switches the gene off. Because 5’-CG-3 is palindromic, C on one or both strands may be methylated (hemi- or homo-methylated respectively). DNA methylation is mediated by methyl-CpG-binding proteins. DNA is wrapped around histone proteins to which acetyl groups can be attached which switches the gene on (acetylation) or removed which switches the gene off (deacetylation). Seven mammalian sirtuin proteins act as NAD-dependent deacetylases for genome stability. Histones may also be modified by methylation. Small forms of RNA (RNAi) can interfere with gene expression by attaching to other RNA molecules. Many cancers exhibit hypomethylation of oncogenes switching them on and hypermethylation of tumour suppressor genes switching them off (Balmain, Reference Balmain1995). The silencing of tumour suppressor genes through methylation is as effective as genetic mutation but unlike mutation are reversible (Plass and Soloway, Reference Plass and Soloway2002).

Role of viruses

Transposable elements are DNA sequences that can be cut-and-paste from and into the genome (jumping genes). They are classed into two types: (i) DNA transposons were active during early primate evolution until 37 My ago since when they have been inactive; (ii) retrotransposons form RNA intermediates which are then reverse transcribed into DNA for insertion into the chromosome as (a) long terminal repeats (LTR) or human endogenous retroviruses (HERV) and (b) non-LTR including long interspersed elements (LINE), Alu or SVA elements. Around 45% of the human genome comprises transposable elements which have causative roles in human rapid evolution, genetic instability and cancer (Ayarpadikannan and Kim, Reference Ayarpadikannan and Kim2014). Around 8% of the human genome comprises fossil endogenous retroviral (ERV) genes in addition to other random insertions of retrotransposons such as Alu and LINE elements. Evolution of ERVs is dependent on the high error rate of reverse transcriptase and recombination through its jumps (Blikstad et al., Reference Blikstad, Benachenhou, Sperber and Blomberg2008). Retroviruses comprise RNA genomes from which a double-strand DNA copy is constructed through reverse transcriptase and the viral DNA is integrated into the host chromosomal DNA using viral integrase enzyme forming a provirus ~7–11 kb in length (Temin, Reference Temin1972). Retroviruses generally integrate at random locations in host chromosomes but have preference for active locations – others such as ALV have distinct target preferences. Integration of retroviral DNA into the host chromosome enhances the expression of adjacent downstream host genes – if the activated gene is an oncogene, tumour growth may result. Retroviruses may have simple genomes (alpha, beta, gamma and epsilon) or complex genomes (lentivirus, deltavirus and spumavirus) but only simple retroviruses have become endogenous in their hosts. Endogenous avian leukosis virus (ALV) is an alpha-retrovirus that infects chickens similar to Rous sarcoma virus (RSV) inducing B-cell lymphomas – ALV (5’LTR-gag-pol-env-LTR3’) does not transform fibroblasts but RSV (5’LTR-gag-pol-src-LTR3’) does (Weiss, Reference Weiss2013). The order of structural genes is conserved in all retroviruses: gag gene encodes for virion structure; pol encodes for reverse transcriptase and other enzymes and env encodes envelope proteins for host cell recognition. Env and gag markers were inherited as a single host locus. Each of the LTRs comprises unique U3 and U5 regions separated by an R segment. The transformation-capable RSV genome is 20% greater than the transformation-defective genome due to the inclusion of a cancer-causing src gene near the 3’ end. Endogenous retroviruses have evolved from retroviral infections in the ancestral germline which subsequently became fixed and then vertically transmitted from parent to offspring. This has been occurring since retroviruses emerged 500 My ago. The env sequence of a HERV provirus on chromosome 7q21.2 has been co-opted to express protein Syncytin-1 in the placenta (Jern and Coffin, Reference Jern and Coffin2008). There are around 30 human ERV (HERV) families in the human genome, most of which have accumulated mutations so they do not offer uninterrupted open reading frames. Of the 4000 endoviral env loci in the human genome, only 16 retain intact open reading frames that encode an Env protein (De Parseval et al., Reference De Parseval, Lazar, Casella, Benit and Heidmann2003). HERV gag and pol have even lower ratios of 17/9500 and 13/20 900 loci respectively. There are no known replicative HERV in the human genome including HERV-K(HML-2) which is the most active HERV family and was provirally integrated some 200 000 years ago. ERVs have been implicated in tumourigenesis including HERV, an infectious example of which is human T cell leukaemia virus 1 (HTLV-1) (Ruprecht et al., Reference Ruprecht, Mayer, Sauter, Roemer and Mueller-Lantzsch2008). Oncogenes of retroviruses are copies of host cellular proto-oncogenes that are implicated in carcinogenesis (Bishop, Reference Bishop1983). ERV can induce cancer through two mechanisms: (i) oncogene capture involves infection by transforming retrovirus rapidly causing tumour growth due to the expression of a viral oncogene (v-onc) derived from host cell proto-oncogenes (c-onc) that were inserted into the retroviral genome through recombination during reverse transcription; (ii) insertional mutagenesis involves the integration of proviral genomes close to host cell proto-oncogenes that activate them or into tumour-suppressor genes which are disrupted and inactivated allowing slower tumour growth in both cases. Immunosuppressive domains are present in 10 of the 16 HERV elements containing complete env genes in the human genome suggesting that HERV act as tumour antigens that elicit anti-tumour immune response. ERV also jump between species (xenotropism) – RD114 is a feline ERV found in the tabby but not spotted cats that was infected from a primate as it is found in baboons (Weiss, Reference Weiss2006). There are well-known exogenous viruses that are associated with human cancers. Human tumour-inducing viruses yield prolonged latency to malignant growth and include Epstein-Barr virus (EBV), hepatitis B virus (HBV), hepatitis C virus (HCV), human papilloma virus (HPV), human T cell lymphotropic virus (HTLV-1) and Kaposi's sarcoma virus (KSV) constituting 15–20% of human cancers globally (Javier and Butel, Reference Javier and Butel2008; Martin and Gutkind, Reference Martin and Gutkind2009). Oncolytic viruses preferentially lyse cancer cells by activating dormant tumours to activate immune response but operate best in combination with other therapies (Melcher et al., Reference Melcher, Harrington and Vile2021). Cancer forms a co-evolving ecosystem in which cancer therapies can eradicate most cancer cells but also promote clonal evolution of drug resistance (Ushijima et al., Reference Ushijima, Clark and Tan2021). Interferon, a type of cytokine regulatory transcription factor, is the first viral defence in mammals that blocks the spread of viral infection through the body by invoking cell death of infected cells (Sen, Reference Sen2001). Some use double-stranded RNA as cofactor that activates promoters for their activity against RNA viruses. They act through different cell surface receptors. Most viruses have evolved defences to maintain a dynamic equilibrium between interferon and virus. Oncolytic viruses are viruses that attack tumours, e.g. the modified adenovirus Onyx-015 targets cancer cells that have mutated p53 genes (in its unmodified form, it possesses genes that inactivates p53). They can work with checkpoint inhibitors such as ipilimumab. Rapamycin is a TOR signalling network inhibitor as a treatment for malignant tumours. A chimaeric antigen receptor has been engineered to modify a T lymphocyte to act as an OR gate in recognizing two cancer antigens – CD19 or CD20 – to prevent escape by mutation.

Biomimetic lessons in counting

It has been proposed that there is a tiny erosion of mean telomere length per generation of any species that very gradually leads to critically short telomeres and genomic instability in the germline inducing species extinction (species clock hypothesis) (Stindl, Reference Stindl2004). Such a process would be too short for a self-replicating machine. Alternatively, an ageing process may be envisaged – in biological organisms, the accumulation of spontaneous somatic mutations is the primary mechanism of ageing (Curtis, Reference Curtis1963). Post-reproductive lifespans extending the onset of senescence will evolve only when the offspring and other relatives benefit (Hamilton, Reference Hamilton1966). There are two primary theories of ageing (Partridge and Barton, Reference Partridge and Barton1993; Rando, Reference Rando2006). The pleiotropic theory of ageing predicts that genes that are advantageous in early life are selected even if they are disadvantageous in later life (Williams, Reference Williams1957; Bell Reference Bell1984) due to the decline of selection pressure with age particularly after reproductive maturation (Charlesworth, Reference Charlesworth2000), e.g. the onset of Huntington's disease occurs after reproductive prime. Cancer is a direct – though not universal – result of the pleiotropy theory of ageing (Campisi, Reference Campisi2005). The mechanism of pleiotropic theory of ageing may be that telomeric shortening inhibits cancer formation in early life but simultaneously involves progressive degradation with age (reserve capacity hypothesis) (Weinstein and Ciszek, Reference Weinstein and Ciszek2002). Consistent with the pleiotropic theory of ageing is the disposable soma theory that increased longevity imposes metabolic costs for the maintenance of somatic tissue against deterioration (Kirkwood and Rose, Reference Kirkwood and Rose1991). Both may be incorporated into a single stochastic dynamic programming model based on the Gompertz ageing equations (Mangel, Reference Mangel2001). Reliability theory of ageing predicts exponential mortality rates with age (Gompertz law) in biological systems rather than power law mortality rates with age (Weibull law) in engineered systems (Gavrilov and Gavrilova, Reference Gavrilov and Gavrilova2001). Biological systems' extensive adoption of multiple, overlapping functional redundancies renders an exponential profile that begins with infant mortality failures and increases rather than the bathtub power law profile of high infant-reliability of engineered components. The Gompertz model estimates the maximum human lifespan at 120 years – death grows exponentially with age until a plateau at 105–113 years, following which there is a sharp growth in mortality beyond 113. By implementing high-reliability manufacture, test and recycling of components, a self-replicating machine can exhibit a Weibull power law ageing profile.

If evolutionary selection is not operational in a self-replicating machine, senescence may be imposed by energetic limitations to controlling the growth of errors in information-based molecules (Kirkwood, Reference Kirkwood1977). It has been demonstrated in animal models that treatment with the non-integrative adeno-associated virus expressing TERT to activate telomerase yielded improved health and longevity without increasing the incidence of cancer (De Jesus et al., Reference De Jesus, Vera, Schneeberger, Tejera, Ayuso, Bosch and Blasco2012; De Jesus and Blasco, Reference De Jesus and Blasco2013). We must ensure that there is no such happy valley between senescence and cancer for our self-replicating machine.

Biomimetic lessons into engineering design

Our first consideration concerns the physical implementation of our genetic instruction store for a self-replicating machine. Based on the material resources available on the Moon (Ellery, Reference Ellery2020), we propose that magnetic core memory constitutes the most appropriate medium of information storage. Magnetic core memories were used for RAM over the period 1955–75 and are tolerant of radiation with high reliability – they were used in the Apollo Guidance Computer and Space Shuttle Flight Computers. The chief advantage of magnetic core memory over solid-state electronics is its much lower susceptibility to radiation errors increasing the reliability of data storage. Magnetic core memory has many of the basic materials and components required in a dc electric motor (Elaskri and Ellery, Reference Elaskri and Ellery2018). There are several magnetic core materials. Soft iron loses its magnetization when the field is removed. It can concentrate fields up to 2 T but suffers from eddy currents due to its good electrical conductivity generating waste heat. Laminated silicon steel reduces its electrical conductivity and so hinders eddy current formation. Powdered iron coated in a thin layer of insulation is an alternative. The core is typically constructed from ferrite (Fe3O4) ceramic through which wires are passed to convey read and write signals. Each core stores one bit of information non-volatilely as zero or one depending on the direction of the core's magnetization. The toroidal shape of the core allows close packing of the cores. Tiny magnetic toroids are formed in the cores forming arrays of cores. The unit cell of magnetic core memory comprises a toroidal hysteresis loop of magnetically hard material to act as a transformer core material (high remanence but low coercivity material such as cobalt ferrite or AlNiCo) through which several (aluminium) wires pass for reading/writing non-volatile binary data. Electric current passing through the wires set or reset the bit stored magnetically in the core. The cores are magnetized clockwise or anticlockwise to represent binary 1 and 0 or vice versa. A set of three wires (X, Y, Sense/Inhibit) through the centre of the core carries current to generate a magnetic field concentrated by the core. A grid of horizontal (x) and vertical (y) wires generate the write current while two perpendicular diagonal wires are the read wires. Only at the junction where the x and y wires cross (logical AND) is there sufficient current to magnetically write a bit. Any change in magnetic state induces a voltage pulse in a read line which in turn invokes a re-write. The invention of the coincident current system enabled a small number of wires to control a large number of cores in 3D stacks. A large number of small ferrite toroidal cores are held on layers of XY grids of wires through the toroidal centres. Only where the combined magnetic field from X and Y lines cross exceeds a threshold will the magnetic polarity reverse. Although non-volatile, reading data resets the core so a re-write cycle is required to restore the data after reading. In the 1960s, the limit of memory storage density was ~1 kbit l−1, i.e. 1 Mbit m−3. Although there exist today a suite of manufacturing technologies that may increase this density, this is currently conjectural. If we assume that the genetic program will be invariant, the program can be woven into read-only core rope memory which is renowned for its high reliability in space applications. Binary data was stored either as wire threaded through the core (1) or around the core (0) which was read as a current pulse. This offers a higher storage density of 2.5 Mbit m−3 than that of magnetic core memory with higher reliability.

We propose that the physical design of the genetic memory module of the self-replicating machine comprises a volumetric array of magnetic core memory cells. Copying of uninterpreted genetic instructions from a parent self-replicating machine to an offspring self-replicating machine is a fundamental requirement of self-replication (von Neumann and Burks, Reference von Neumann and Burks1966). Memory copying from one storage location to another is a fundamental capability in computing, e.g. memcpy in C. We assume arbitrarily that a copying head scans the original magnetic core memory array serially and/or in parallel to copy it onto an identical blank magnetic core memory array. It is important to note that biological template copying is of a different nature to artificial copying – in the dividing cell, the parent DNA unwinds in which both daughter cells each retain one original parent DNA strand from which to re-constitute the double strand. In effect, the parent no longer exists – during this process of replication, telomeres in both daughter cells are shortened. Any further replication cycles will undergo further shortening in offspring. In artificial self-replication, the parent and offspring are distinct entities. In order for analogous telomeres to act as counters, the parent telomere must be truncated as well as those of the offspring copies. We suggest that analogous to biological telomeres, a physical linear tail of blank memory cells feeds into the original to-be-read memory array, the number of cells corresponding to the maximum number of replications (Hayflick limit). For example, to limit the population of self-replicating machines to 1.5 million, the number of replications (generations) is limited to 13 (assuming two offspring per generation) by P = (1 + r)i where P = population, r = number of offspring/generation, i = generation number. However, two offspring per generation offers double the population growth rate over a single offspring – a population of 1.5 million requires 13 generations compared with 20–21 generations of single offspring. Greater than two offspring per generation offer diminishing returns in terms of rapidity – three offspring per generation requires over 10 generations to yield a population of 1.5 million. For Galactic colonization, 24 generations yield a population of 424 billion machines assuming two offspring per generation, i.e. the maximum number of offspring is constructed by the first generation of probes at 48.

Both parent and any offspring must exhibit telomeres that are reduced by Ci +1 = Ci-1 – physical reduction provides greater robustness than logical reduction implying that a blank memory cell of the tail is physically removed as the initial step in self-replication of any offspring. This must be repeated for every subsequent generation of offspring manufactured by the original parent. The magnetic core memory array of each generation of offspring is copied from the parent in four stages applied at each generation: (i) parent tail counter is shortened by a single memory cell; (ii) a second counter controls the subsequent processes in determining the number of offspring per generation (nominally two); (iii) a blank hardware substrate is assembled including the reduced tail counter; (iv) data copying from parent memory to offspring memory. As each offspring inherits a reduced counter tail, subsequent offspring will generate further generations with telomeric tails in lockstep with the original parent. This system will be reliable for a single offspring every generation. The implementation of two offspring with identical counters per generation requires an additional counting loop to be incorporated such as a toggle switch. This introduces an additional mechanism and a concomitant greater chance of error including the prospect of uncontrolled replication if the additional counter malfunctions. This is exactly analogous to biological cancer in which telomerase genetic expression malfunctions leading to uncontrolled proliferation.

We can also implement computer viruses stored in magnetic core memory that are instantiated under undesirable conditions through physical interrupts to infect the neural network architectures such as if the outer counter exceeds two offspring per generation. Computer malware is a family of malicious computer programs that include spyware, ransomware, adware, rootkits, Trojan horses, worms and viruses. Computer viruses are executive software programs designed to interfere with computer operations that replicates itself through computer networks (Cohen, Reference Cohen1987; Denning, Reference Denning1988). Unlike a worm which is an independent self-replicating program, it cannot exist independently of a host program. The Trojan horse part of the virus is the functional code that implements the payload of the virus by writing its own code into the host program. The payload may implement cyber-attacks such as physical attack on hardware devices and their drivers, e.g. SCADA; reconnaissance attacks searching for vulnerabilities through traffic analysis, denial-of-service, access attacks through network ports and privacy attacks. The virus is triggered as a logic bomb that activates it once an event or condition occurs. In our case, this event is the initiation of a third circuit around the number of offspring per generation loop. To minimize the size of the viral payload, the payload must cripple a crucial capability systematically. We decided to specifically attack the computational system.

Our computational system is based on analogue neural networks comprising networks of neuron modules based on a modified version of the Yamashita-Nakamura neuron (Yamashita and Nakamura, Reference Yamashita and Nakamura2007; Prasad and Ellery, Reference Prasad and Ellery2020). Specifically, the neural network architecture configurations of fixed (pre-trained) connection weights are stored in magnetic core memory. One relatively straightforward approach would be to initialize all stored connection weights to zero. This would then print all analogue neural network circuits with severed synaptic links to ensure that all computational circuits are defunct. However, metal whisker growth is a well-known but poorly understood phenomenon which can occur under a variety of environmental conditions which could potentially grow undesired conductive paths between the severed analogue neurons. Alternatively, a quine is a self-referencing program that copies its own code. It is implemented by a virus that modifies memory. The neural network quine is any neural network that learns to output its own weights on inputs defining the coordinates of the connection weight, i.e. self-replicating (Chang and Lipson, Reference Chang and Lipson2018). The loss function is defined as the difference between the weight and the predicted weight output: $L = \sum _i\lfloor{y_i-w_i} \rfloor ^2$. Although it is possible for the self-replicating neural network to perform auxiliary tasks, this degrades its ability to self-replicate. Hence, training for self-replication will degrade the network's ability to perform other functions which is desirable. However, the network must not be input with predicted weight outputs as this would iteratively shrink the weights towards zero, i.e. severed connection links which introduces the aforementioned problem. Nevertheless, the implementation of a virus in magnetic memory that reconfigures all neural network weights to a quine configuration would effectively disable self-replicating machines' ability to procreate.

Conclusions

We have examined the biology of telomeres in implementing the Hayflick limit to biological cellular replication. Reliable imposition of a Hayflick limit will be essential in preventing uncontrolled replication of self-replicating machines. We propose a physical telomeric counter that is operational prior to the copying of genetic instructions that limits the generation number from a parent machine and any subsequent offspring. However, introducing the flexibility of multiple offspring per generation opens up the prospect of cancerous replication cycles. This illustrates the necessity for multiple layers of mechanisms to enhance reliability to prevent uncontrolled replication. A final safety layer would be a stored virus invoked under certain conditions such as replicating more than two offspring per generation that propagates through memory to initialize all neural network circuits with weights that output their own values, i.e. in effect, viral induction of brain cancer.

Conflict of interest

There is no conflict of interest associated with this paper.

References

Aisner, D, Wright, W and Shay, J (2002) Telomerase regulation: not just flipping the switch. Current Opinion in Genetics & Development 12, 8085.CrossRefGoogle ScholarPubMed
Ait-Aissa, K, Ebben, J and Kadlec, A (2016) Friend or foe? Telomerase as a pharmacological agent in cancer and cardiovascular disease. Pharmacological Research 111, 422433.CrossRefGoogle ScholarPubMed
Arkus, N (2005) Mathematical model of cellular apoptosis and senescence through the dynamics of telomere loss. Journal of Theoretical Biology 235, 1332.CrossRefGoogle ScholarPubMed
Armanios, M and Blackburn, E (2012) Telomere syndromes. Nature Review Genetics 13, 693704.CrossRefGoogle ScholarPubMed
Austad, S and Fischer, K (1991) Mammalian aging, metabolism and ecology: evidence from the bats and marsupials. The Journals of Gerontology 46, B47B53.CrossRefGoogle ScholarPubMed
Autexier, C and Greider, C (1996) Telomerase and cancer: revisiting the telomere hypothesis. Trends in Biochemical Sciences 21, 387391.CrossRefGoogle ScholarPubMed
Ayarpadikannan, S and Kim, H-S (2014) Impact of transposable elements in genome evolution and genetic instability and their implications in various diseases. Genomics Information 12, 98104.CrossRefGoogle ScholarPubMed
Balmain, A (1995) Exploring the bowels of DNA methylation. Current Biology 5, 10131016.CrossRefGoogle ScholarPubMed
Bar, P (1996) Apoptosis – the cell's silent exit. Life Sciences 59, 369378.CrossRefGoogle ScholarPubMed
Bekaert, S, Derradj, H and Baatout, S (2004) Telomere biology in mammalian germ cells and during development. Developmental Biology 274, 1530.CrossRefGoogle ScholarPubMed
Bell, G (1984) Evolutionary and nonevolutionary theories of senescence. American Naturalist 124, 600603.CrossRefGoogle Scholar
Ben-Porath, I and Weinberg, R (2005) Signals and pathways activating cellular senescence. The International Journal of Biochemistry & Cell Biology 37, 961976.CrossRefGoogle ScholarPubMed
Bensaad, K and Vousden, K (2007) P53: new roles in metabolism. Trends in Cell Biology 17, 286291.CrossRefGoogle ScholarPubMed
Bishop, M (1983) Cellular oncogenes and retroviruses. Annual Reviews Biochemistry 52, 301354.CrossRefGoogle ScholarPubMed
Blackburn, E (1992) Telomerases. Annual Reviews Biochemistry 61, 113129.CrossRefGoogle ScholarPubMed
Blasco, M and Hahn, W (2003) Evolving views of telomerase and cancer. Trends in Cell Biology 13, 289294.CrossRefGoogle ScholarPubMed
Blikstad, V, Benachenhou, F, Sperber, G and Blomberg, J (2008) Evolution of human endogenous retroviral sequences: a conceptual account. Cellular & Molecular Life Science 65, 33483365.CrossRefGoogle ScholarPubMed
Bodnar, A, Ouellette, M, Froklis, M, Holt, S, Chiu, C-P, Morin, G, Harley, C, Shay, J, Lichsteiner, S and Wright, W (1998) Extension of life-span by introduction of telomerase into normal human cells. Science 279, 349352.CrossRefGoogle ScholarPubMed
Bortner, C, Oldenburg, N and Cidlowski, J (1995) Role of DNA fragmentation in apoptosis. Trends in Cell Biology 5, 2126.CrossRefGoogle ScholarPubMed
Bryan, T, Englezou, A, Dalla-Pozza, L, Dunham, M and Reddel, R (1997) Evidence for an alternative mechanism for maintaining telomere length in human tumours and tumour-derived cell lines. Nature Medicine 3, 12711274.CrossRefGoogle Scholar
Cahill, D, Kinzler, K, Vogel stein, B and Lengauer, C (1999) Genetic instability and Darwinian selection in tumours. Trends in Genetics 15, M57M60.CrossRefGoogle Scholar
Campisi, J (2001) Cellular senescence as a tumour-suppressor mechanism. Trends in Cell Biology 11, S27S31.CrossRefGoogle Scholar
Campisi, J (2005) Aging, tumour suppression and cancer: high wire act!. Mechanisms of Aging & Development 126, 5158.CrossRefGoogle ScholarPubMed
Chakhparonian, M and Wellinger, R (2003) Telomere maintenance and DNA replication: how closely are these two connected? Trends in Genetics 19, 439446.CrossRefGoogle ScholarPubMed
Chang, O and Lipson, H (2018) Neural network quine. arXiv:1803.05859v4 [cs.AI] 24 May 2018.CrossRefGoogle Scholar
Charlesworth, B (2000) Fisher, Medawar, Hamilton and the evolution of aging. Genetics 156, 927931.CrossRefGoogle ScholarPubMed
Chiu, C-P and Harley, C (1997) Replicative senescence and cell immortality: the role of telomeres and telomerase. Proceedings of the Society for Experimental Biology and Medicine 214, 99106.CrossRefGoogle ScholarPubMed
Cohen, F (1987) Computer viruses: theory and experiments. Computer & Security 6, 2235.CrossRefGoogle Scholar
Collins, K (2000) Mammalian telomeres and telomerase. Current Opinion in Cell Biology 12, 378383.CrossRefGoogle ScholarPubMed
Crespi, B and Summers, K (2005) Evolutionary biology of cancer. Trends in Ecology & Evolution 20, 545549.CrossRefGoogle ScholarPubMed
Curtis, H (1963) Biological mechanisms underlying the aging process. Science 141, 686694.CrossRefGoogle ScholarPubMed
Danial, N and Korsmeyer, S (2004) Cell death: critical control points. Cell 116, 205219.CrossRefGoogle ScholarPubMed
De Jesus, B and Blasco, M (2013) Telomerase at the intersection of cancer and aging. Trends in Genetics 29, 513520.CrossRefGoogle Scholar
De Jesus, B, Vera, E, Schneeberger, K, Tejera, A, Ayuso, E, Bosch, F and Blasco, M (2012) Telomerase gene therapy in adult and old mice delays aging and increases longevity without increasing cancer. EMBO Molecular Medicine 4, 691704.CrossRefGoogle Scholar
De Parseval, N, Lazar, V, Casella, J-F, Benit, L and Heidmann, T (2003) Survey of human genes of retroviral origin: identification and transcriptome of the genes with coding capacity for complete envelope proteins. Journal of Virology 77, 1041410422.CrossRefGoogle ScholarPubMed
Denning, D (1988) Computer viruses. American Scientist 76, 236238.Google Scholar
Dubrana, K, Perrod, S and Gasser, S (2001) Turning telomeres off and on. Current Opinion in Cell Biology 13, 281289.CrossRefGoogle ScholarPubMed
Dunham, M, Neumann, A, Fasching, C and Reddel, R (2000) Telomere maintenance by recombination in human cells. Nature Genetics 26, 447450.CrossRefGoogle ScholarPubMed
Eisenstein, M (2022) Restoring The Guardian of the genome. Nature 603, S1S5.CrossRefGoogle Scholar
Elaskri, A and Ellery, A (2018) Developing techniques to 3D print electric motors. Proc Int Symp Artificial Intelligence Robotics & Automation in Space, Madrid, Spain, paper no. 10c-2.Google Scholar
Elaskri, A and Ellery, A (2020) 3D printed electric motors as a step towards self-replicating machines. Proc Int Symp Artificial Intelligence, Robotics and Automation in Space, paper no 5020.Google Scholar
Ellery, A (2015 a) Engineering artificial extraterrestrial life? Proc European Conf on Artificial Life (Late Breaking), York, UK, 12–14.Google Scholar
Ellery, A (2016) Are self-replicating machines feasible? AIAA Journal of Spacecraft and Rockets 53, 317327.CrossRefGoogle Scholar
Ellery, A (2017 a) Building physical self-replicating machines. Proc European Conf on Artificial Life, Lyon, France, 146–153.CrossRefGoogle Scholar
Ellery, A (2017 b) Space exploration through self-replication technology compensates for discounting in NPV cost-benefit analysis – a business case? New Space Journal 5, 141154.CrossRefGoogle Scholar
Ellery, A (2020) Sustainable in-situ resource utilisation on the moon. Planetary & Space Science 184, 104870.CrossRefGoogle Scholar
Ellery, A (2021) Are there biomimetic lessons from genetic regulatory networks for developing a lunar industrial ecology? Biomimetics Journal 6, 50.CrossRefGoogle ScholarPubMed
Ellery, A and Eiben, G (2019) To evolve or not to evolve: that is the question. Proc Artificial Life Conf, 357–364.Google Scholar
Ellery, A, Mellor, I, Wanjara, P and Conti, M (2022) Metalysis FFC process as a strategic lunar in-situ resource utilisation technology. New Space Journal 10, 224238.CrossRefGoogle Scholar
Epel, E, Blackburn, E, Lin, J, Dhabhar, F, Adler, N, Morrow, J and Cawthorn, R (2004) Accelerated telomere shortening in response to life stress. Proceedings of the National Academy of Sciences of the United States of America 101, 1731217315.CrossRefGoogle ScholarPubMed
Fadeel, B and Orrenius, S (2005) Apoptosis: a basic biological phenomenon with wide-ranging implications in human disease. Journal of Internal Medicine 258, 479517.CrossRefGoogle ScholarPubMed
Finkel, T, Serrano, M and Biasco, M (2007) Common biology of cancer and ageing. Nature 448, 767774.CrossRefGoogle ScholarPubMed
Flores, I and Blasco, M (2010) Role of telomeres and telomerase in stem cell aging. FEMS Letters 584, 38263830.CrossRefGoogle ScholarPubMed
Freitas, R (1980) Self-reproducing interstellar probe. The Journal of the British Interplanetary Society 33, 251264.Google Scholar
Freitas, R (2001) Some limits to global ecophagy by biovirus nanoreplicators with public policy recommendations. reprint.Google Scholar
Gavrilov, L and Gavrilova, N (2001) Reliability theory of aging and longevity. Journal of Theoretical Biology 213, 527545.CrossRefGoogle ScholarPubMed
Griffith, J, Comeau, L, Rosenfeld, S, Stansei, R, Bianchi, A, Moss, H and de Lange, T (1999) Mammalian telomeres end in a large duplex loop. Cell 97, 503514.CrossRefGoogle Scholar
Gupta, S (2001) Molecular steps of death receptor and mitochondrial pathways of apoptosis. Life Sciences 69, 29572964.CrossRefGoogle ScholarPubMed
Hamilton, W (1966) Moulding of senescence by natural selection. Journal of Theoretical Biology 12, 1245.CrossRefGoogle ScholarPubMed
Hayflick, L (1979) Cell biology of aging. Journal of Investigative Dermatology 73, 814.CrossRefGoogle ScholarPubMed
Hayflick, L (1984) Intracellular determinants of cell aging. Mechanisms of Aging & Development 28, 177185.CrossRefGoogle ScholarPubMed
Hoeijmakers, J (2001) Genome maintenance mechanisms for preventing cancer. Nature 411, 366374.CrossRefGoogle ScholarPubMed
Javier, R and Butel, J (2008) History of tumour virology. Cancer Research 68, 76937706.CrossRefGoogle Scholar
Jern, P and Coffin, J (2008) Effects of retroviruses on host genome function. Annual Review Genetics 42, 709732.CrossRefGoogle ScholarPubMed
Joy, B (2000) Why the future doesn't need us. Wired (Apr). Available at https://www.wired.com/2000/04/joy-2/.Google Scholar
Kahneman, D and Tversky, A (1979) Prospect theory: an analysis of decision under risk. Econometrica 47, 263291.CrossRefGoogle Scholar
Kass-Eisler, A and Greider, C (2000) Recombination in telomere-length maintenance. Trends in Biochemical Sciences 25, 200204.CrossRefGoogle ScholarPubMed
Kelleher, C, Teixerira, T, Forstemann, K and Lingner, J (2002) Telomerase: biochemical considerations for enzyme and substrate. Trends in Biochemical Sciences 27, 572579.CrossRefGoogle ScholarPubMed
Kierszenbaum, A (2000) Telomeres: more than chromosomal non-sticking ends. Molecular Reproduction & Development 57, 23.3.0.CO;2-R>CrossRefGoogle ScholarPubMed
Kim, N, Piatyszek, M, Prowse, K, Harley, C, West, M, Ho, P, Coviello, G, Wight, W, Weinrich, S and Shay, J (1994) Specific association of human telomerase activity with immortal cells and cancer. Science (New York, N.Y.) 266, 20112014.CrossRefGoogle ScholarPubMed
Kirkwood, T (1977) Evolution of aging. Nature 270, 301304.CrossRefGoogle Scholar
Kirkwood, T and Rose, M (1991) Evolution of senescence: late survival sacrificed for reproduction. Philosophical Trans Biological Sciences 332, 1524.Google ScholarPubMed
Krantic, S, Mechawar, N, Reix, S and Quirion, R (2006) Molecular basis of programmed cell death involved in neurodegeneration. Trends in Neuroscience 28, 670676.CrossRefGoogle Scholar
Kruk, P, Rampino, N and Bohr, V (1995) DNA Damage and repair in telomeres: relation to aging. Proceedings of the National Academy of Sciences of the United States of America 92, 258262.CrossRefGoogle ScholarPubMed
Land, H, Parada, L and Weinberg, R (1983) Cellular oncogenes and multistep carcinogenesis. Science 222, 771778.CrossRefGoogle ScholarPubMed
Lansdorp, P (2005) Major cutbacks at chromosome ends. Trends in Biochemical Sciences 30, 388395.CrossRefGoogle ScholarPubMed
Lentzos, F, Rybicki, E, Engelhard, M, Paterson, P, Sandholtz, A and Reeves, G (2022) Eroding norms over release of self-spreading viruses. Science 375, 3133.CrossRefGoogle ScholarPubMed
Lundblad, V (2001) Genome instability: McClintock revisited. Current Biology 11, R957R960.CrossRefGoogle ScholarPubMed
Mangel, M (2001) Complex adaptive systems, aging and longevity. Journal of Theoretical Biology 213, 559571.CrossRefGoogle ScholarPubMed
Martin, D and Gutkind, J (2009) Human tumour-associated viruses and new insights into the molecular mechanisms of cancer. Oncogene 27, S31S42.CrossRefGoogle Scholar
Masutomi, K, Yu, E, Khurts, S, Ben-Porath, I, Currier, J, Metz, G, Brooks, M, Kaneko, S, Murakami, S, DeCaprio, J, Weinberg, R, Stewart, S and Hahn, W (2003) Telomerase maintains telomere structure in normal human cells. Cell 114, 241253.CrossRefGoogle ScholarPubMed
McCormick, F (1999) Signalling networks that cause cancer. Trends in Cell Biology 9, M53M56.CrossRefGoogle ScholarPubMed
McEachern, M, Krauskopf, A and Blackburn, E (2000) Telomeres and their control. Annual Reviews Genetics 34, 331358.CrossRefGoogle ScholarPubMed
Melcher, A, Harrington, K and Vile, R (2021) Oncolytic virotherapy as immunotherapy. Science (New York, N.Y.) 374, 13251326.CrossRefGoogle ScholarPubMed
Moyzis, R, Buckingham, J, Cram, S, Dani, M, Deaven, L, Jones, M, Meyne, J, Ratliff, R and Wu, J-R (1988) Highly conserved repetitive DNA sequence (TTAGGG)n present at the telomeres of human chromosomes. Proceedings of the National Academy of Sciences of the United States of America 85, 6622–2226.CrossRefGoogle ScholarPubMed
Munoz-Espin, D, Canamero, M, Maraver, A, Gomez-Lopez, G, Contreras, J, Murillo-Cuesta, S, Rodriguez-Baeza, A, Varela-Nieto, I, Ruberte, J, Collado, M and Serrano, M (2013) Programmed cell senescence during mammalian embryonic development. Cell 155, 11041118.CrossRefGoogle ScholarPubMed
Nagley, P and Wei, Y-H (1998) Ageing and mammalian mitochondrial genetics. Trends in Genetics 14, 513517.CrossRefGoogle ScholarPubMed
Nakamura, T and Cech, T (1998) Reversing time: origin of telomerase. Cell 92, 587590.CrossRefGoogle ScholarPubMed
Neidle, S and Parkinson, G (2003) Structure of telomeric DNA. Current Opinion in Structural Biology 13, 275283.CrossRefGoogle ScholarPubMed
Olivier, M, Hollstein, M and Hainaut, P (2010) TP53 Mutations in human cancers: origons, consequences and clinical use. Cold Spring Harbor Perspectives in Biology 2, a001008.CrossRefGoogle Scholar
Olovnikov, A (1996) Telomeres, telomerase and aging: origin of the theory. Experimental Gerontology 31, 443448.CrossRefGoogle ScholarPubMed
Olshansky, J, Carnes, B and Desesquelles, A (2001) Prospects for human longevity. Science 291, 14911492.CrossRefGoogle ScholarPubMed
Partridge, L and Barton, N (1993) Optimality, mutation and the evolution of aging. Nature 362, 305311.CrossRefGoogle Scholar
Peltomaki, P (1997) DNA Mismatch repair gene mutations in human cancers. Environmental Health Perspectives 105, 775780.Google Scholar
Plass, C and Soloway, P (2002) DNA Methylation, imprinting and cancer. European Journal of Human Genetics 10, 616.CrossRefGoogle ScholarPubMed
Prasad, V and Ellery, A (2020) Analogue neural network architecture for in-situ resourced computing hardware on the Moon. Proc Int Symp Artificial Intelligence, Robotics and Automation in Space, paper no 5005.Google Scholar
Pray, L (2008) DNA Replication and causes of mutation. Nature Education 1, 214218.Google Scholar
Radman, M, Matic, I, Halliday, J and Taddei, F (1995) Editing DNA replication and recombination by mismatch repair: from bacterial genetics to mechanisms of predisposition to cancer in humans. Philosophical Trans Royal Society London B 347, 97103.Google ScholarPubMed
Rando, T (2006) Stem cells, aging and the quest for immortality. Nature 441, 10801086.CrossRefGoogle ScholarPubMed
Ruprecht, K, Mayer, J, Sauter, M, Roemer, K and Mueller-Lantzsch, N (2008) Endogenous retroviruses and cancer. Cellular & Molecular Life Science 65, 33663382.CrossRefGoogle ScholarPubMed
Sagan, C and Newman, W (1983) Solipsist approach to extraterrestrial intelligence. Quarterly Journal of the Royal Astronomical Society 24, 113121.Google Scholar
Sahin, E and DePinho, R (2010) Linking functional decline of telomeres, mitochondria and stem cells during ageing. Nature 464, 520528.CrossRefGoogle ScholarPubMed
Schumacher, B, Pothof, J, Vijg, J and Hoeijmakers, J (2021) Central role of DNA damage in the ageing process. Nature 592, 695703.CrossRefGoogle ScholarPubMed
Sen, G (2001) Viruses and interferons. Annual Review Microbiology 55, 255281.CrossRefGoogle ScholarPubMed
Sherr, C (2004) Principles of tumour suppression. Cell 116, 235246.CrossRefGoogle Scholar
Simonsson, T (2003) Substrate for telomerase. Trends in Biochemical Sciences 28, 632638.CrossRefGoogle ScholarPubMed
Soto, A and Sonnenschein, C (2004) Somatic mutation theory of cancer: growing problems with the paradigm? BioEssays 26, 10971107.CrossRefGoogle ScholarPubMed
Stindl, R (2004) Is telomere erosion a mechanism of species extinction? The Journal of Experimental Zoology Part B: Molecular and Developmental Evolution 302B, 111120.CrossRefGoogle Scholar
Stone, M (2018) Closer look at human telomerase. Nature 557, 174175.CrossRefGoogle Scholar
Temin, H (1972) RNA tumour viruses – background and foreground. Proceedings of the National Academy of Sciences of the United States of America 69, 10161020.CrossRefGoogle ScholarPubMed
Tubbs, A and Nussenzweig, A (2017) Endogenous DNA damage as a source of genomic instability in cancer. Cell 168, 644656.CrossRefGoogle ScholarPubMed
Turner, L (2004) Biotechnology, bioethics and anti-aging interventions. Trends in Biotechnology 22, 219221.CrossRefGoogle ScholarPubMed
Ushijima, T, Clark, S and Tan, P (2021) Mapping genomic and epigenomic evolution in cancer ecosystems. Science 373, 14741478.CrossRefGoogle ScholarPubMed
Venkatesan, S, Khaw, K and Hande, P (2017) Telomere biology – insights into an intriguing phenomenon. Cells 6, 6020015.CrossRefGoogle ScholarPubMed
Vixtorelli, S and Passos, J (2017) Telomeres and cell senescence – size matters not. EBioMedicine 21, 1420.CrossRefGoogle Scholar
von Neumann, J and Burks, A (1966) Theory of Self-Reproducing Automata. Champaign, Illinois: University of Illinois Press.Google Scholar
Weinstein, B and Ciszek, D (2002) Reserve capacity hypothesis: evolutionary origins and modern implications of the trade-off between tumor-suppression and tissue repair. Experimental Gerontology 37, 615627.CrossRefGoogle ScholarPubMed
Weiss, R (2006) Discovery of endogenous retroviruses. Retrovirology 3, paper number 67.CrossRefGoogle ScholarPubMed
Weiss, R (2013) On the concept and elucidation of endogenous retroviruses. Philosophical Transactions of the Royal Society B368, 201220494.Google Scholar
White, L, Wright, W and Shay, J (2001) Telomerase inhibitors. Trends in Biotechnology 19, 114120.CrossRefGoogle ScholarPubMed
Williams, G (1957) Pleiotropy, natural selection and the evolution of senescence. Evolution 11, 398411.CrossRefGoogle Scholar
Wodaerz, D and Komarova, N (2007) Can loss of apoptosis protect against cancer? Trends in Genetics 23, 232237.CrossRefGoogle Scholar
Wright, W and Shay, J (1995) Time, telomeres and tumours: is cellular senescence more than an anticancer mechanism? Trends in Cell Biology 5, 293297.CrossRefGoogle ScholarPubMed
Wright, W, Piatyszek, M, Rainey, W, Byrd, W and Shay, J (1996) Telomerase activity in human germline and embryonic tissues and cells. Developmental Genetics 18, 173179.3.0.CO;2-3>CrossRefGoogle ScholarPubMed
Xi, H, Li, C, Ren, F, Zhang, H and Zhang, L (2013) Telomere, aging and age-related diseases. Aging Clinical Experimental Research 25, 139146.CrossRefGoogle ScholarPubMed
Yamashita, Y and Nakamura, Y (2007) Neuron circuit model with smooth nonlinear output function. Proc Int Symp Nonlinear Theory & its Applications, Vancouver, 11–14.Google Scholar
Zhang, J and Xu, M (2002) Apoptotic DNA fragmentation and tissue homeostasis. Trends in Cell Biology 12, 8489.CrossRefGoogle ScholarPubMed