Synthetic Human Chromosomes

Frontier The prize is a delivery vehicle for whole gene programs that no virus can carry: a stand-alone human chromosome that sits beside the 46 a person already has, replicates and segregates with them, and never inserts into the host genome. The single-copy human artificial chromosome reported in 2023 was built from a 760 kb construct and, in a companion demonstration, was engineered to carry on the order of a megabase of added DNA — roughly 150 to 200 times the 4.7 kb of cargo an adeno-associated virus vector can package, the workhorse of in-body gene therapy. That is the specific thing on offer: not a larger edit but a different category of payload, an entire regulated locus with its own introns, enhancers and insulators rather than a stripped complementary DNA.

Established Measured against today’s vectors, the magnitude is a payload jump of two orders and the removal of insertional mutagenesis as a risk. A lentivirus integrates semi-randomly and has caused leukaemia in trial; an artificial chromosome is episomal by construction. The one that works loses at 0.011 per cell per day without any drug selection — about one copy in ninety divisions — which is the number that makes the idea more than a curiosity. Frontier Everything else about it is still contested: de novo formation runs at 42 to 46 per cent of drug-resistant clones under favourable conditions, and it has been shown only in cancer cell lines, not in the primary or stem cells a therapy would use.

Speculative The near-term beneficiary is the laboratory, not the patient, and that distribution matters. A reliable single-copy chromosome is first a research reagent — a place to park a humanised gene cluster, a whole metabolic pathway, or a safe-harbour transgene at defined dose — and its value accrues to whoever can synthesise and verify megabase tandem repeats, a capability that today sits in perhaps a handful of laboratories worldwide. A written chromosome placed in a person is a later and separate proposition.

Handwave The whole upside rests on one thing nobody can yet do: write a centromere. If a human centromere can be specified de novo — built from designed sequence rather than seeded on borrowed natural alpha-satellite and left to an epigenetic lottery — then the payload vehicle above becomes designable rather than found, and the same step is what turns SynHG’s synthetic-chromosome goal from an assembly feat into a working chromosome. That condition, and the segregation test in section 7 that would confirm it, is what the rest of this brief is about.

Established A human chromosome is a linear DNA molecule that must do three mechanical jobs to be inherited: replicate from multiple origins, cap its ends with telomeres so they are not read as breaks, and carry one centromere that assembles a kinetochore to pull it to a daughter cell at each division. Writing one from scratch means supplying all three functions in designed sequence and getting a live human cell to treat the result as a real chromosome. The gene content is the easy part; the machinery that makes it a chromosome rather than a large plasmid is the hard part. Established Of the three jobs, two are sequence problems that current methods already handle — origins come along with almost any large human DNA, and telomeres are short defined repeats — and one, the centromere, is not.

Established Two routes exist, and they are almost opposites. The top-down route takes a real human chromosome, ships it into another cell — historically a rodent cell, by microcell-mediated chromosome transfer — and truncates and edits it down, keeping the natural centromere intact. The bottom-up route introduces naked centromeric DNA into a human cell and lets a de novo centromere and a de novo chromosome self-assemble. The bottom-up product is what most people mean by a human artificial chromosome (HAC); the top-down product is really an engineered natural chromosome. A written chromosome in the strict sense — sequence designed end to end, including the part that segregates it — has never been made. Frontier Most of what the field calls progress lives in the gap between these two routes, and reading which route a given result used is the first thing to check about any claim of a “synthetic” human chromosome.

Established The reason the substrate is different from bacterial or yeast genome writing is the centromere, and the difference is roughly four orders of magnitude. A budding-yeast centromere is a sequence-defined point centromere about 125 base pairs long; the entire genome is about 12 megabases. A human centromere is a regional, epigenetically specified structure sitting on megabases of near-identical alpha-satellite repeat, which alone makes up 2.8 per cent of the genome. Fabricating a yeast chromosome, which the Sc2.0 consortium has now done sixteen times, leaves the centromere problem essentially untouched, because a 125 bp cassette can simply be synthesised and dropped in. That is why organism-scale synthetic biology, covered in Synthetic Biology, does not answer this question. Established The regulatory object is different too: a bacterial or yeast genome is a contained industrial construct, while a human chromosome is, from the first design choice, a candidate for inheritance in a person.

Frontier The field’s central paradox is that the centromere is defined epigenetically but forms de novo only on a specific sequence. Which chromatin becomes a centromere is marked not by DNA sequence but by nucleosomes containing the histone variant CENP-A, which is why neocentromeres can appear on ordinary DNA lacking alpha-satellite and why the sequence under a human centromere can vary. Yet in the laboratory a new centromere assembles reliably only when the input DNA is alpha-satellite carrying CENP-B protein binding sites (“CENP-B boxes”). The maintenance is sequence-independent; the establishment is not. Nobody can yet start an epigenetic centromere on arbitrary designed sequence to order.

Frontier The programme that has made this an explicit goal is SynHG. In June 2025 the Wellcome Trust committed 10 million pounds over five years to the Synthetic Human Genome project, led by Jason Chin (MRC Laboratory of Molecular Biology and the Generative Biology Institute at the Ellison Institute of Technology, Oxford), with partners at Cambridge, Kent, Manchester and Imperial College London. Its stated first target is the toolkit to synthesise a single human chromosome — about 2 per cent of a person’s DNA — as a proof of concept within five to ten years. Speculative Its own framing is that a complete synthetic human genome, and any clinical use, is a matter of decades.

Established The reference against which any synthetic chromosome is now measured is the telomere-to-telomere assembly of a complete human genome, finished in 2022, which for the first time resolved the centromeres. The T2T-CHM13 assembly showed that satellite repeats make up 6.2 per cent of the genome and that alpha-satellite, the centromeric repeat, is the single largest component at 2.8 per cent. Before this, centromeric DNA could not be assembled at all — the arrays are too large and too repetitive for short-read sequencing — so the thing a synthetic chromosome most needs to reproduce was, until four years ago, literally unreadable. Established Reading a centromere is strictly easier than writing one, and it took until 2022; the ordering of those two milestones is the single most useful fact for calibrating how far off a written chromosome is.

Established The same work pinned down how the active centromere is positioned, and it is not by sequence. Within each alpha-satellite array the most recently expanded repeats are the ones most likely to bind CENP-A, the inner-kinetochore histone, and these coincide with a local dip in DNA methylation — the “centromere dip region.” The kinetochore sits on an epigenetically chosen sub-region of a megabase array, and which sub-region can move. This is the empirical basis for calling the human centromere epigenetic, and it is the fact a de novo design has to contend with.

Established De novo human artificial chromosomes have existed since 1997, when introduced alpha-satellite arrays were first shown to nucleate new centromeres and self-assemble into microchromosomes. The founding results (Harrington and colleagues; Ikeno and colleagues) established the recipe that has held ever since: supply enough alpha-satellite carrying CENP-B boxes, deliver it to a permissive human cell line — almost always the fibrosarcoma line HT1080 — and a fraction of clones will carry a new mitotically stable chromosome. The recipe works; it has never been efficient, controllable, or portable to the cells that matter. Established That HT1080 is the near-universal host is itself a warning sign: a recipe that works in one permissive line and few others is a phenomenon, not yet a method.

Frontier How inefficient is the quantity that has held the field back, and it is measured. Introduced synthetic alphoid-tetO constructs form HACs about 5 per cent of the time, against about 30 per cent for natural alpha-satellite clones. Worse, formation is accompanied by uncontrolled multimerisation: the input BAC or YAC amplifies into megabase-sized arrays, so the product has an unpredictable size and copy number even when it works. For thirty years a HAC has been something you select for and characterise after the fact, not something you specify. Frontier Combining CENP-B-box-positive and -negative synthetic repeats has pushed formation from about 4 per cent toward the natural ~30 per cent in one study, the clearest sign that the efficiency gap is a design problem rather than a hard limit.

Frontier The most important recent change is the 2023 demonstration of single-copy HAC formation, which attacked the multimerisation problem by refusing to fight it. Rather than optimise a small construct against its tendency to amplify, the group (Gambogi, Kixmoeller and colleagues, with the Black laboratory at the University of Pennsylvania and collaborators at JCVI and Edinburgh) delivered a 760 kb construct large enough to build the two chromatin domains a working centromere needs on its own, bypassing multimerisation entirely. The result was present in a single copy per cell. Prior Penn HACs of around 200 kb had made many copies of themselves after installation, which is exactly what defeats dose control. Frontier The insight that a larger input is more controllable than a smaller one, because it can build both the open centromeric and closed heterochromatic domains without amplifying, inverts thirty years of instinct and is the conceptual core of the result.

Frontier The single-copy work also reports the numbers that decide whether the object is usable. Across two recipient lines, 42 per cent (SD 9) and 46 per cent (SD 5) of drug-resistant clones carried the HAC — a de novo establishment efficiency above all earlier versions — and the average daily loss rate without antibiotic selection was 0.011 (SD 0.006) over 30 days. The recipients were HT1080 and the osteosarcoma line U2OS, both engineered with an inducible mCherry-LacI-HJURP tether to seed CENP-A. Speculative Whether the same efficiency survives the move to primary or induced-pluripotent stem cells, which do not tolerate the same tricks, is unmeasured and is the open question that matters most.

Established The proof that a whole eukaryotic genome can be fabricated is now in hand, and it is yeast, not human. The Sc2.0 consortium, running since 2006, has synthesised and debugged all sixteen chromosomes of Saccharomyces cerevisiae plus a new-to-nature transfer-RNA neochromosome; the final chromosome, synXVI, a 902,994 bp redesign of the 948,066 bp original, was reported complete in January 2025. Frontier The parts are done; consolidating all sixteen synthetic chromosomes into one living strain is not yet reported finished, and the highest published figure is a strain more than 50 per cent synthetic assembled by breeding. Yeast is the existence proof for design-and-build at genome scale, and it is silent on the human centromere. Established Its real lesson is procedural: synXVI shipped with defects at named loci that had to be found and repaired one at a time, which is what “debugging” a written chromosome actually looks like.

Established The blunt position statement is the anchor of this brief: as of September 2026 no synthetic human chromosome exists. What exists is a complete map of the thing to be copied (T2T), a controllable single-copy de novo HAC in cancer cell lines, a finished synthetic yeast genome, and a five-year, 10-million-pound programme that has named the synthetic human chromosome as a goal and stated it will probably not reach it. The gap between those assets and the goal is the subject of the rest of this brief. Frontier The most defensible one-line summary of the position is that the field has learned to read the hardest part of a chromosome and to control the copy number of an artificial one, but not to write the part that makes a chromosome segregate.

Frontier Can a de novo centromere be started on designed, non-natural sequence? Every reliable HAC to date is seeded on alpha-satellite with CENP-B boxes. Whether a fully engineered sequence — the point of “writing” a chromosome — can nucleate CENP-A chromatin without borrowing natural centromeric DNA is unresolved. The alphoid-tetO system and CENP-A tethering experiments show a centromere can be forced open at a chosen array, but forcing is not the same as specifying from sequence alone. Speculative If the answer is no — if natural alpha-satellite carries information no design captures — then “writing” a human chromosome may always mean copying a real centromere, which would be a smaller and more honest claim than the field’s rhetoric.

Frontier Why does synthetic alpha-satellite underperform natural alpha-satellite by roughly six-fold? The measured 5 per cent versus 30 per cent gap between alphoid-tetO and natural clones is unexplained at the mechanistic level. Candidate factors — the density and phasing of CENP-B boxes, higher-order repeat organisation, the exact array length — are being manipulated one at a time, and combining CENP-B-box-positive and -negative repeats has pushed formation from about 4 per cent toward the natural ~30 per cent in one study, but no design rule predicts the outcome in advance.

Frontier Does single-copy formation transfer out of transformed cell lines? HT1080 and U2OS are permissive in ways primary cells are not; they tolerate transfection, have altered checkpoints, and in U2OS maintain telomeres without telomerase. The entire clinical case depends on the same object forming and persisting in induced-pluripotent stem cells or primary haematopoietic cells, and there is no published efficiency figure there. Frontier This is the cheapest decisive experiment available — it needs no new construct, only a harder recipient — which is why its absence from the literature is itself informative.

Speculative Can megabase tandem repeats be synthesised and verified at all, at acceptable cost? Writing a chromosome the honest way means writing the alpha-satellite array, which is megabases of near-identical repeat — the hardest possible substrate for chemical synthesis, hierarchical assembly and sequence verification, because near-identical units collapse and misassemble. No published series prices synthesis at this scale, and it is not obvious the current oligo-to-assembly pipeline reaches it without a new method. Frontier One escape is to keep borrowing natural alpha-satellite and write only the rest, but that concedes the centromere is copied rather than designed, and the field has not decided whether that counts as a synthetic chromosome.

Frontier How is a whole chromosome delivered into a target cell without a rodent intermediate? Microcell-mediated chromosome transfer is inefficient and passes the chromosome through a mouse cell, where, as the SynHG teams note, human telomeres lengthen and then revert on return. Alternatives — direct large-DNA transfection, whole-nucleus or synthetic-cell delivery — are unproven at chromosome scale. Delivery is a bottleneck the bottom-up route has never had to solve because it builds in place.

Speculative What is the minimal specification of a functional human chromosome? Nobody knows the smallest working combination of origin density, telomere length, centromeric array size and gene content, nor whether a designed chromosome needs cohesin-loading and replication-timing features that natural chromosomes have and current HACs may lack. The Sc2.0 experience — where redesign introduced defects that had to be found and fixed one locus at a time — suggests the specification is discovered, not deduced. Speculative A human chromosome has far more such features than a yeast chromosome and far less tolerance for getting them wrong, so the discovery process is likely to be longer, not shorter, than yeast’s two decades.

Established The single hardest bottleneck is that the centromere is epigenetic, so it cannot be specified by sequence the way every other chromosomal element can. Origins, telomeres and genes are sequence features that can be written and will work. The centromere is a chromatin state that has to be induced and then inherited, and the only reliable way to induce it uses natural alpha-satellite. Until de novo CENP-A establishment on designed sequence is solved, a “written” human chromosome still contains a piece that was found, not written.

Frontier Megabase tandem-repeat synthesis and verification is a manufacturing bottleneck with no published solution. Alpha-satellite arrays are the worst case for every step: synthesis (repeats cause polymerase slippage and hairpins), assembly (near-identical units recombine and delete), and verification (short reads cannot span the array, and even long reads struggle to place near-identical monomers). This is why the T2T project needed new sequencing to read centromeres at all; writing them is strictly harder than reading them. Speculative It is possible that no purely chemical route to a megabase array exists and that the only path is enzymatic or cell-based amplification of a smaller seed, which would make the “written” centromere partly grown rather than synthesised.

Frontier Copy-number and size control has only just become tractable and only in permissive cells. The thirty-year multimerisation problem — input DNA amplifying into unpredictable megabase arrays — was the reason a HAC could not be a dosed reagent. The 760 kb single-copy result shows the problem is solvable by going large, but a large construct reintroduces the synthesis and delivery bottlenecks above, so the bottlenecks trade against each other rather than falling together.

Frontier Delivery into therapeutically relevant cells is unsolved. Microcell-mediated chromosome transfer works in the laboratory at low efficiency and depends on a rodent host cell; it is not a route to a human therapy. No method reliably moves a megabase-scale chromosome into a primary human cell at useful yield, and the cells that would receive it — stem cells, T cells — are exactly the ones that resist large-DNA delivery.

Speculative Cost is an unmeasured bottleneck that could dominate. Oligonucleotide synthesis is cheap per base, but assembled, error-corrected, chromosome-scale DNA is not priced anywhere, and the repetitive fraction may be uneconomic by any current method. A chromosome that can be built once as a demonstration but not manufactured is a scientific result, not a platform, and the field has no number that settles which it is.

Frontier Long-term mitotic and meiotic stability at the standard a therapy needs is unproven. A loss rate of 0.011 per day over 30 days is good for a research reagent; a therapy delivered to a self-renewing stem-cell pool needs stability over years and hundreds of divisions, and behaviour in meiosis — relevant to any germline concern — is essentially uncharacterised for HACs.

Established This topic depends first on the ability to write and assemble large DNA, which it consumes rather than produces. Every synthetic chromosome is downstream of the DNA-writing capability owned by Genetic Engineering and the organism-scale assembly demonstrated by Synthetic Biology. The distinct dependency here is assembly of the repetitive fraction, which those fields have largely been able to avoid because coding genes and regulatory DNA are not tandem repeats.

Established It depends on a complete, accurate sequence of what it is copying, which arrived only in 2022. The T2T reference is the design template: without a base-level map of alpha-satellite higher-order repeats and the location of the CENP-A domain, there was no specification to write to. The dependency runs one way — reading had to be finished before writing could be attempted. Established The reference is also drawn from a single, near-haploid source, so the design template encodes one genome’s centromeres, not the human range of them.

Frontier It depends on centromere biology that is still being worked out. The rules for de novo CENP-A establishment, the role of CENP-B and its boxes, and why synthetic arrays underperform natural ones are active research questions, not settled inputs. A design discipline cannot exist ahead of the mechanism, and the mechanism is a frontier. Speculative It is not guaranteed that the mechanism, once known, will be specifiable in sequence at all — some epigenetic states may be reachable only through a particular history, not a particular sequence.

Established It depends on long-read sequencing to verify what was built. A synthetic alpha-satellite array can only be confirmed correct with sequencing that spans it; the same technology that let T2T read centromeres is a hard prerequisite for validating any written centromere, and its error modes on near-identical repeats set the ceiling on how confidently a build can be called correct. Frontier A build that cannot be verified is a build that cannot be trusted as a therapeutic, so verification is a gating dependency, not a downstream convenience.

Speculative It depends on a delivery method that does not yet exist for therapy. The bottom-up route sidesteps delivery by building in place, but building in place in a patient is not on offer; a therapeutic version depends on a whole-chromosome delivery method that is currently a research bottleneck rather than an available tool. Frontier The top-down route’s reliance on a rodent intermediate is the clearest illustration that delivery, not design, may be the step that keeps this out of the clinic longest.

Frontier The decisive experiment is a de novo synthetic human chromosome segregating faithfully through 50 or more divisions in a primary human cell line without an intact donated centromere. Each clause is load-bearing. “Synthetic” means the sequence is written, not a truncated natural chromosome. “Without an intact donated centromere” means the centromere is established on designed input rather than seeded on a natural alpha-satellite array lifted from a real chromosome. “Primary human cell line” rules out the permissive transformed lines where every current result lives. “50 or more divisions” sets a stability bar a research reagent has cleared but a written chromosome in a normal cell has not. This is the single result that would move this brief from “never been done” to “done.”

Frontier A necessary precursor is de novo centromere establishment on fully designed sequence, with no natural alpha-satellite. A cleaner version tests only the centromere: deliver an engineered array — sequence chosen by design rules, not copied — and measure whether CENP-A chromatin assembles and a kinetochore forms. A positive result would show the epigenetic centromere can be specified from scratch; a negative result across the plausible design space would be the strongest evidence that natural alpha-satellite is doing something no design yet captures.

Frontier The transfer test: repeat single-copy HAC formation in induced-pluripotent stem cells and report the efficiency. This needs no new construct, only the recipient the clinical case actually requires. A formation rate anywhere near the 42 to 46 per cent seen in HT1080 and U2OS would be a major result; a rate near zero would relocate the field’s optimism to where the evidence is.

Speculative The synthesis test: build and verify a megabase synthetic alpha-satellite array and prove it is correct by long-read sequencing. Before any chromosome can be written, someone has to show the repetitive fraction can be manufactured and confirmed at scale. This is a pure engineering milestone with a clear pass/fail, and no published result reaches it. Frontier It is also separable from the biology: an array can be built and verified without ever asking a cell to form a centromere on it, which makes it the one experiment here that a manufacturing group could run today.

Frontier The dose test: place a defined single-copy payload on a HAC and show constant expression and constant copy number across many divisions. The whole advantage over integrating vectors is dose control and no insertional mutagenesis; that advantage is a claim until someone shows expression that does not drift as an integrating construct’s would. The 0.011-per-day loss figure is the closest existing measurement and it is for the chromosome’s presence, not for stable expression of a payload.

Speculative The delivery test: move a formed chromosome into a primary human cell without a rodent intermediate. A demonstration that a megabase chromosome can be delivered directly, at a reported yield, would remove the one step the bottom-up route has never solved and the top-down route solves only through mouse cells. Its absence is why every therapeutic timeline in section 13 is long.

Handwave The programme-level test is whether SynHG delivers a synthetic human chromosome, or reports foundational tools instead, inside its five-year grant. The programme’s own language makes the second outcome the likely one; treating the tools milestone as the honest deliverable, and the chromosome as the aspiration, is the reading this brief takes. Which one lands by 2030 is itself an informative result about how hard the substrate is.

Established The engineering breaks into four elements that must all work in one molecule: origins, telomeres, a centromere, and payload. Three of the four are, individually, solved or nearly so. Replication origins are dense in the human genome and come along with almost any large human DNA. Telomeres are short defined repeats that can be added synthetically. Payload — genes with their regulatory context — is exactly what large-DNA assembly does well. The centromere is the element that is not an engineering problem in the ordinary sense, because it is a chromatin state rather than a sequence. Frontier The practical consequence is that three-quarters of the parts list can be ordered and the last quarter cannot, so build effort concentrates entirely on the one element that resists specification.

Frontier Construct size is now a design variable, not just a constraint. The single-copy result inverted the old instinct to keep constructs small: a 760 kb input that can build both the open (centromeric) and closed (heterochromatic) domains on its own avoids multimerisation, so bigger is more controllable. The engineering cost is that a larger construct is harder to synthesise, clone and deliver, which is why the recommendation to carry at least a megabase of DNA is simultaneously a solution and a new problem.

Established Assembly is hierarchical and borrows from yeast. Large synthetic constructs are built by transformation-associated recombination and stepwise assembly in yeast and bacterial artificial chromosomes, the same toolchain that built the Sc2.0 chromosomes in Synthetic Biology. What does not transfer is the repetitive centromeric array, which recombines and deletes in exactly the hosts used to assemble everything else; a repeat-tolerant assembly method is the missing tool. Speculative Whether generative design or new assembly chemistry supplies that tool first is unknown, and it is the kind of enabling advance that would move several of this brief’s bottlenecks at once.

Frontier Verification is a first-class engineering requirement, not an afterthought. A build is only as trustworthy as the sequencing that confirms it, and near-identical repeats are where sequencing is least trustworthy. The engineering standard a written centromere has to meet — and nobody has set it — is a long-read validation that can distinguish a correct array from a subtly collapsed one.

Speculative Delivery hardware is the least developed element. For research, building in place removes the problem; for anything downstream, moving a chromosome requires either improved microcell transfer or a genuinely new large-cargo method. This is where the topic touches synthetic-cell and vector engineering, and where it is furthest from a working tool.

Established Long-read sequencing is the enabling adjacent technology, and it arrived just in time. Nanopore and high-fidelity long-read platforms are what made the T2T centromere assembly possible and are the only way to verify a synthetic array. Progress in reading repeats sets the pace at which anyone can confidently write them.

Established Large-DNA assembly and the yeast toolchain are shared with organism-scale synthetic biology. The methods that built synthetic yeast chromosomes — hierarchical assembly, SCRaMbLE-style recombination, error-corrected synthesis — are the same ones a human chromosome build would use for its non-repetitive parts, which is why Synthetic Biology and Genetic Engineering are the immediate neighbours.

Frontier Generative sequence design is being enlisted, with unproven reach. SynHG’s partners describe using generative AI and robotic assembly; the same protein- and DNA-design methods that produce binders and regulatory elements in Synthetic Biology are being pointed at chromosome design. Whether a model can propose an alpha-satellite array that forms a centromere is exactly the open frontier question, so the adjacency is a hope, not yet a tool. Speculative A generative model trained on the T2T centromere maps could in principle propose arrays to test, but there is no published case of a designed centromeric sequence forming a centromere, so the method has not yet earned the claim.

Speculative Cell and gene therapy is the adjacent application pull. A single-copy, non-integrating, high-capacity vehicle is attractive wherever a whole locus or pathway must be delivered, and the topic borders Genetic Engineering, Cellular Rejuvenation and Biological Computing, where large synthetic circuits would benefit from a chromosome-scale host. In each case the adjacency is downstream of a chromosome that does not yet exist. Frontier The nearest real application is not therapy but engineered cell lines, where a single-copy chromosome-scale host would let a whole synthetic pathway be carried at defined dose without competing for space in the native genome.

Frontier The defining institutional fact is that a single funder has made this an explicit, named programme with a deliverable and a clock. Wellcome’s 10-million-pound, five-year commitment to SynHG in 2025 concentrated the field around one goal — a synthetic human chromosome — and one honest caveat, that it will probably not be reached in the grant. That is a different institutional object from the diffuse HAC literature that preceded it.

Established The programme built its ethics and social-science stream in from the start rather than bolting it on. A parallel initiative, Care-full Synthesis, led by Joy Zhang at the University of Kent, runs across the whole grant and includes empirical work in Africa, Asia and the Americas on genomic representation. Whether integration changes outcomes or supplies legitimacy is untested, but the structural choice is deliberate and unusual.

Frontier There is no regulatory object here yet, because there is no product. A synthetic human chromosome placed in a person would be a first-of-kind for every regulator: neither a small edit, nor an integrating vector, nor a cell therapy, but an added chromosome. No pathway exists for evaluating an object whose whole point is to be inherited by every daughter cell, and building that pathway is an institutional prerequisite this brief flags as unmet.

Speculative The capability is concentrated in a handful of laboratories, which is both a governance lever and a fragility. Megabase repeat assembly and de novo HAC formation are practised in very few places; that makes oversight tractable and progress slow and hostage to a few groups. Concentration is easy to govern and easy to stall. Frontier It also means that the honest current answer to “who can build this?” is a list of named laboratories rather than an industry, which is the clearest single indicator of how immature the field is.

Frontier DNA-synthesis screening is the institutional edge case that touches this topic sideways. Chromosome-scale synthesis passes through commercial and benchtop synthesis whose screening regime is contested — the 2024 US framework for nucleic-acid synthesis screening was rescinded in 2025, as recorded in Synthetic Biology. A programme writing human sequence at scale sharpens a screening question the field had not resolved. Speculative A chromosome-writing capability is not obviously a pathogen-synthesis capability — the hard parts here, centromeres and delivery, are not what a bad actor needs — so the sharpest biosecurity concern may be diffuse rather than specific to this topic.

Frontier The load-bearing ethical distinction is somatic versus germline, and a written chromosome makes it unusually sharp. A chromosome designed to segregate faithfully is, by construction, heritable in any cell that undergoes meiosis; the same property that makes it a good vehicle in a stem cell makes germline transmission a design feature rather than an accident. Nothing about the technology enforces the somatic-only line, which is why the line has to be institutional.

Speculative The public framing runs immediately to designer babies and synthetic humans, and the researchers have engaged it rather than dismissed it. One senior scientist has publicly noted the capability “could one day be used to create synthetic humans, biological weapons, or even creatures with human DNA,” while stressing the distance; a funder has argued the work is safer done openly now than left to others later. These are the honest poles of the debate and neither is settled.

Frontier Dual-use is real but not unique to this topic. The ability to synthesise long human sequence is the same ability biosecurity worries about generally; a chromosome programme concentrates it but does not create it, and the meaningful safeguards are the synthesis-screening ones that sit upstream of every project, discussed in Synthetic Biology.

Established Representation is a concrete, addressable equity problem the programme has named. Reference genomes and centromere maps are drawn from few individuals — the T2T assembly came from a single haploid source — so any design discipline built on them risks encoding a narrow slice of human variation. The empirical social-science work embedded in the programme is aimed squarely at this, which makes it one of the few ethical issues here with a defined action rather than only a worry.

Speculative Enhancement is a distant question that the science does not yet license. Adding a designed chromosome carrying enhancing gene sets is conceivable in principle and impossible in practice; treating it as imminent misstates where the difficulty lies, which is the centromere and the delivery, not the payload. The ethics that will actually bite first are about first-in-human safety of an inherited object, not about enhancement. Frontier Keeping the debate calibrated to where the difficulty is — the centromere and the delivery — is itself an ethical act, because a debate anchored to enhancement misdirects both scrutiny and reassurance.

Speculative If a human chromosome becomes designable, the unit of biological engineering changes from the gene to the chromosome, and that is the civilizational hinge. Today engineering means editing what is there; a written chromosome means adding a self-inherited, high-capacity module that a cell treats as its own. The consequence is not a faster version of gene therapy but a new place to put biology — whole pathways, safe-harbour capacity, orthogonal genetic systems — that does not compete with the existing genome for space.

Speculative The most likely first civilizational effect is industrial, not medical. A dosed, non-integrating chromosome in engineered cell lines would upgrade biomanufacturing and research before it reaches a clinic, much as synthetic yeast is being framed as a manufacturing platform rather than a medical one. The gains would accrue first to whoever controls the synthesis and assembly capacity, a concentration this brief has already flagged. Speculative If the industrial use arrives and the medical use does not, the lasting civilizational effect of this line of work could be a better bioreactor rather than a new medicine — a smaller prize than the headlines, and a real one.

Handwave A capability to write chromosomes is a capability to write inheritable human change, and civilizations have no settled way to hold that. The germline property is intrinsic, the governance is not built, and the technology’s own advocates place the hard version decades out. The sober reading is that the civilizational stakes are large and the timeline long enough that institutions could be built ahead of the capability — if the will exists, which is the handwave.

Speculative The downside distribution is the usual one for a rare, expensive capability: it widens rather than narrows access gaps unless deliberately countered. A chromosome-scale therapy, if it ever arrives, would be among the most expensive interventions conceivable, and the reference maps it is built on encode a narrow ancestry. Left alone, the benefits concentrate; the programme’s equity stream is a bet that this can be steered, and it is only a bet. Handwave The optimistic case — that building the capability openly and slowly now lets institutions mature alongside it — is the funder’s stated rationale and is, at this stage, an assertion about human institutions rather than a result about the technology.

These horizons track construction and stability milestones for a written human chromosome, not a technology curve, because the only curve that exists is a price curve for the input and a reading curve for the reference:

• 10 yr: Frontier Single-copy HAC formation demonstrated in induced-pluripotent stem cells with a reported efficiency is the most likely landmark, and it needs no new construct. Frontier SynHG most plausibly reports foundational tools — assembly, delivery, verification — rather than a finished chromosome, consistent with the tools covered in Genetic Engineering and Synthetic Biology. Speculative A verified megabase synthetic alpha-satellite array is possible but unclaimed.
• 25 yr: Speculative A de novo centromere established on fully designed, non-natural sequence — the precursor the decisive experiment needs — is a plausible mid-horizon result if the mechanism yields. Speculative A written human chromosome segregating through 50-plus divisions in a primary line is at the edge of this window at best.
• 50 yr: Speculative A synthetic chromosome used as a dosed research and biomanufacturing reagent at industrial scale, well ahead of any clinical use. Handwave A first-in-human somatic chromosome-scale therapy, contingent on a delivery method and a regulatory pathway that do not exist today.
• 100 / 250+ yr: Handwave Chromosome-scale design as a routine part of human medicine or biology, and any germline application, sit beyond the horizon this brief can honestly resolve; the technology’s own advocates place the hard version in decades and decline to place the hardest version at all.

• Depends on Genetic Engineering, which owns DNA writing and editing as a general capability, and Synthetic Biology, which owns organism-scale fabrication and the yeast assembly toolchain. This topic consumes both and adds one substrate they largely avoid: the repetitive centromeric array. It also waits on the telomere-to-telomere reference that finished in 2022, without which there was no specification to write to.
• Requires (not on this map) First, a scientific result nobody is producing: a de novo centromere that assembles CENP-A chromatin on designed sequence rather than on natural alpha-satellite with CENP-B boxes. Second, an industrial capacity to synthesise and verify megabase tandem repeats, the worst-case substrate for every synthesis, assembly and sequencing step, which no published series prices or demonstrates. Third, error-corrected assembly priced at chromosome scale, a manufacturing capability a supplier could choose to build but has not. Fourth, a way to deliver a whole megabase chromosome into a primary human cell without a rodent intermediate, currently a research bottleneck rather than a tool. Fifth, an institutional result: a regulatory pathway for evaluating an added, inheritable human chromosome, which no regulator has, because no such product exists.
• Enables Single-copy, non-integrating, high-capacity delivery of whole loci and pathways for cell engineering and, much later, therapy; a chromosome-scale host for the large synthetic circuits of Biological Computing; and orthogonal genetic systems in Xenobiology. No typed enabling edge is claimed, because in each case what is enabled waits on a chromosome that does not yet exist.
• Adjacent Synthetic Biology, Genetic Engineering, Designer Organisms, Xenobiology, Biological Computing and Cellular Rejuvenation; and outside the map, long-read sequencing, centromere and kinetochore biology, and gene-therapy vector engineering.

Established “Sc2.0 finished a synthetic genome, so a synthetic human chromosome is next.” The claim in circulation treats yeast as a stepping stone. It is not, for the one part that matters: a yeast centromere is a ~125 bp sequence cassette and a human centromere is a megabase epigenetic structure. Sc2.0 is a genuine landmark for design-and-build and says almost nothing about the human centromere, which is the whole difficulty here.

Frontier “Human artificial chromosomes already exist, so this is mostly done.” HACs have existed since 1997, but they are seeded on natural alpha-satellite, form unpredictably — historically about 5 per cent of the time for synthetic arrays — and until 2023 multimerised into uncontrolled copy numbers. What exists is a found object you select for, not a written chromosome you specify. The gap between the two is this brief’s subject.

Frontier “The single-copy result solved the problem.” It solved copy-number control in permissive cancer cell lines by using a 760 kb construct, which is real and important. It did not write a centromere de novo, did not work in primary or stem cells, and traded the multimerisation problem for a synthesis-and-delivery problem. It is a decisive step on one axis, not the finish line.

Speculative “SynHG will build a synthetic human chromosome in five years.” The programme’s own framing is that its five-to-ten-year target is a toolkit, that a complete chromosome is unlikely within the grant, and that a whole genome and any clinical use are decades away. Reporting the aspiration as the deliverable inverts what the programme actually said.

Speculative “This means designer babies are around the corner.” The germline concern is legitimate because heritability is intrinsic to a chromosome, but the binding constraints are the centromere, the delivery and first-in-human safety of an inherited object — not the payload. The scenario that dominates public discussion is gated behind the problems this brief flags as unsolved.

Handwave “Falling DNA-synthesis cost makes chromosome writing just a matter of time.” Cost per base has fallen steeply, but the binding cost is assembled, error-corrected, repeat-heavy DNA, which nobody prices, plus a centromere that no amount of cheap synthesis specifies. A curve that is real for oligonucleotides is being extrapolated onto a problem it does not address. Established The same category error appears whenever the recoded E. coli and synthetic yeast results are cited as evidence that a human chromosome is near: those are triumphs of writing sequence, and the human chromosome’s hard part is not sequence.