A Cell Is Not a Type: What Evolutionary History Reveals About Identity

What if the most important fact about a cell is not what it is doing now, but what it has been capable of becoming for hundreds of millions of years?

That question sounds biological, but it reaches far beyond biology. We routinely classify the world by visible resemblance: a neuron is a neuron, a muscle cell is a muscle cell, and a sympathetic neuron is a sympathetic neuron. Modern single cell methods have complicated this picture. They reveal enormous variation within familiar categories, while comparisons across species show that cells can preserve deep molecular similarities even when their bodies, environments, and developmental histories differ.

The resulting tension is easy to state: how can a cell remain recognizably the same while changing so much?

The answer is that cellular identity is not a fixed label. It is a negotiated relationship between ancestry, developmental potential, molecular machinery, and present circumstance. Once we see identity this way, cross species analysis and evolutionary developmental biology stop being separate enterprises. Together, they offer a more precise model of how living systems preserve function without freezing form.

The mistake of treating cell types as nouns

A conventional cell atlas resembles a dictionary. It assigns names to clusters of cells according to the genes they express, then treats those names as if they were natural kinds. This is useful, but it can also mislead. A cluster is a measurement outcome, not necessarily a biological boundary.

Consider a city map. One map might divide a metropolis by administrative districts, another by subway lines, another by neighborhoods, and another by patterns of daily movement. None is simply the city itself. Each reveals a different structure. A single cell classification works similarly: it highlights one dimension of a living system while hiding others.

This matters especially when comparing species. Two cells may perform the same broad function while expressing different genes because they inhabit different tissues, developmental environments, or physiological conditions. Conversely, two cells may share a molecular signature while serving different roles because they are embedded in different anatomical circuits.

The tempting solution is to search for a universal molecular barcode, a small set of genes that identifies a cell type everywhere. But evolution rarely preserves an entire barcode. It more often preserves a functional architecture: a set of regulatory relationships, signaling capacities, structural constraints, and developmental options that can be implemented through somewhat different molecular parts.

In other words, conservation does not always mean identical ingredients. It can mean conserved organization.

The deepest form of biological similarity is often not sameness of expression, but sameness of possibility.

A cross species comparison therefore has to distinguish at least three things. First, which features are ancient and stable? Second, which features are adaptations to a particular lineage or tissue? Third, which features merely reflect the cell's current state, such as stress, activity, maturation, or exposure to signals?

Without this separation, researchers can make two opposite errors. They can mistake a temporary state for a permanent type, or mistake a divergent implementation for a wholly new invention. Both errors arise from treating a dynamic history as a static label.

Evolution preserves problems before it preserves solutions

The emergence of sympathetic neurons offers a striking way to think about this problem. These neurons are part of the system that coordinates involuntary bodily functions such as heart activity, blood vessel tone, digestion, and rapid responses to threat. Their significance is not merely that they are neurons. They occupy a particular position in a distributed control system, linking the nervous system to organs throughout the body.

Their evolutionary origin raises a more interesting question than “when did this cell type appear?” The deeper question is: what problem did evolution need to solve, and how did an existing developmental population become capable of solving it?

Neural crest cells are especially revealing in this context because they are migratory and developmentally versatile. They move away from the embryonic nervous system and contribute to a wide range of structures. This makes them less like a box labeled “future sympathetic neuron” and more like a mobile design platform. Depending on location and signals, descendants of this population can enter different developmental programs.

That perspective changes how we imagine evolutionary novelty. A new cell type does not necessarily appear from nothing. It may arise when an existing population acquires a new destination, a new regulatory instruction, or a new interaction with surrounding tissues. The raw material is often not a new gene, but a new use of an old developmental capacity.

This is analogous to software evolution. A programming language may already contain functions for communication, timing, and resource allocation. A new application can emerge by recombining those functions in a new architecture. The application is genuinely novel, even though its components are not. In the same way, a specialized neuronal population can evolve through the redeployment of conserved developmental machinery.

The important unit of innovation is therefore not always the gene or the cell. It may be the developmental route that connects an ancestral population to a new physiological role.

This helps explain why evolutionary comparisons can reveal relationships that a snapshot of adult gene expression might obscure. Two adult populations may look different because they have specialized in different contexts. Yet their developmental trajectories may show a shared origin, shared competence, or shared regulatory logic. Conversely, populations that look similar in adulthood may have arrived there through distinct routes.

Origin is not destiny, but it is evidence. Developmental history supplies constraints that remain invisible if we examine only the final state.

The cell type is a trajectory with a memory

A useful synthesis is to treat cell identity as a four layer system.

The first layer is ancestry: where the cell came from and which developmental populations contributed to it. The second is competence: which fates the cell could have adopted under different conditions. The third is implementation: the molecular machinery currently used to perform its role. The fourth is state: what the cell is doing right now in response to activity, environment, or injury.

These layers are related, but they are not interchangeable.

A cell can share ancestry with another cell while having a different adult function. It can share a function with a distant species while using a different implementation. It can share an implementation with a neighboring cell while being in a different physiological state. And it can temporarily resemble another type because both are responding to the same signal.

Most classification errors occur when one layer is used as a proxy for all four.

Imagine two orchestras performing the same symphony. Their instruments may differ, their acoustics may differ, and their interpretation may differ. Yet listeners can still recognize the composition because the relationships among the parts remain stable. A cell type is often more like a composition than a list of instruments. Its identity lies partly in how genes regulate one another, how the cell interacts with its neighbors, and what role it occupies in a larger circuit.

This leads to a practical model: compare cells along axes, not only categories.

One axis measures molecular conservation. Another measures developmental origin. A third measures anatomical position. A fourth measures physiological function. A fifth measures current state. Two cells may be close on one axis and far apart on another. That is not a failure of classification. It is the biological reality the classification should represent.

For example, a sympathetic neuron in one vertebrate lineage may retain a recognizable regulatory program associated with neural crest development and autonomic function, while differing in the precise expression levels of many genes. The differences may reflect adaptation to body size, organ arrangement, temperature, metabolic demand, or the timing of development. Calling the cells either “the same” or “different” is less informative than asking: which dimensions of identity have been conserved, and which have been rewired?

This is also why single cell data should be interpreted as a map of variation rather than a catalogue of isolated objects. Cells occupy neighborhoods in a multidimensional space. Some neighborhoods are separated by sharp boundaries, while others are connected by gradual transitions. Developmental trajectories can pass through regions that do not correspond neatly to adult categories.

A mature cell type may therefore be understood as a stabilized attractor: a region of biological possibility toward which development tends to push cells. Evolution can preserve the attractor while modifying the route, or preserve the route while changing the attractor. This distinction is crucial. Similar endpoints do not guarantee similar histories, and similar histories do not guarantee identical endpoints.

A better way to compare across species

Cross species analysis becomes more powerful when it shifts from the question “Which cells match?” to a sequence of more demanding questions.

First: What function is being compared? A name such as neuron, glia, or immune cell may cover several roles. Functional comparison prevents superficial molecular differences from being overinterpreted.

Second: What developmental origin is being compared? If two populations arise from related embryonic sources, that provides one kind of evidence. If they arise independently but perform similar functions, that suggests convergent evolution.

Third: Which molecular features are stable across context? Genes that persist across developmental stages, tissues, and species are more likely to mark identity than genes that fluctuate with activity or stress.

Fourth: What is the regulatory architecture? Rather than counting shared genes, examine relationships among transcription factors, signaling pathways, and downstream effectors. A conserved network can survive substantial changes in individual components.

Fifth: What does the cell connect to? Anatomy and circuit position can resolve ambiguities left by expression data. A molecularly similar population may have a different role if its targets and inputs differ.

This framework also clarifies what to do with disagreement between datasets. If transcriptomic similarity and developmental origin point in different directions, do not immediately choose one and discard the other. The disagreement may indicate that the population has undergone evolutionary repurposing, or that the measured adult state obscures its developmental history.

The most informative result may be a structured mismatch.

When molecular identity, developmental origin, and function disagree, the disagreement is not noise. It is often the evolutionary story.

This principle has consequences beyond basic research. In regenerative medicine, it suggests that producing a cell with the right marker genes may not be enough. A therapeutic cell may need the appropriate developmental history, connectivity, and response properties. In comparative medicine, it warns against assuming that a cell or disease model transfers perfectly from one species to another merely because a familiar marker is present. In artificial biology, it encourages the design of systems around regulatory relationships and functional constraints rather than isolated parts lists.

Key Takeaways

1. Separate identity from state. When analyzing a cell, distinguish stable developmental and functional features from temporary responses to stress, activity, or environment.

2. Compare architectures, not just markers. Look for conserved regulatory relationships, signaling logic, and circuit position rather than demanding identical gene expression.

3. Treat developmental origin as a second map. A cell's ancestry can reveal relationships that are invisible in its mature molecular profile.

4. Use multidimensional classifications. Record ancestry, competence, implementation, anatomy, function, and current state instead of forcing every cell into a single label.

5. Investigate mismatches. If two forms of evidence disagree, ask what evolutionary repurposing or developmental transition could explain the discrepancy.

The future belongs to historical cell biology

The old image of a cell is spatial: a bounded unit with a recognizable shape and a set of defining parts. The newer image is statistical: a point in a high dimensional expression space. Both are incomplete.

A cell is also historical. It carries traces of an ancestral population, a developmental journey, an evolutionary compromise, and a present negotiation with its environment. Its identity is not simply written in its genes. It is enacted through time.

This reframes the ambition of cell atlases. The goal should not be to produce the final list of all cell types, as though nature had already sorted itself into permanent drawers. The goal is to reconstruct the relationships among origins, states, functions, and possibilities. A useful atlas should tell us not only where a cell is, but where it came from, what it can become, what it is doing, and which aspects of its design are ancient.

The same lesson applies to evolution. Novelty does not require abandoning the past. Evolution is often most creative when it reuses old developmental capacities in new combinations. The emergence of specialized autonomic neurons from a versatile migratory population illustrates a general rule: innovation frequently begins as a change in context, route, or relationship before it becomes a new named thing.

So the next time a biological classification feels uncertain, resist the urge to ask whether the label is correct. Ask a richer question: correct in what sense, and over what timescale?

A cell may be the same by ancestry, different by implementation, similar by function, and unique by state. That is not a contradiction. It is what living identity looks like when viewed honestly: not a noun, but a trajectory with memory.