Optimizing 3D Cell Culture Systems for Human-By-Real Biology Research

The Rise of Human-Relevant Models

For decades, the default place to study human cells has been a flat sheet of plastic. Cells grown on two-dimensional (2D) surfaces are easy to handle and easy to image, and that convenience is real. But one must consider what it costs. On a flat surface, cells take on a morphology, a polarity, and a set of drug responses unlike those in native three-dimensional tissue because they read the stiffness and dimensionality around them and behave to match.1 The deviation from what an intact body would show is not marginal.

To fill that gap, the field has leaned on animal studies for years, and it is worth stripping down what exactly they add. A living animal is an integrated system. It absorbs, distributes, metabolizes, and clears a compound; mounts an immune response; and lets organs signal to one another in a way that only a whole organism can. That whole-body context is precisely what any culture model gives up. At the same time, species differences work against animal models, of course, and contribute to the failure of many compounds that look promising in rodents and then falter in humans, which is itself an argument for alternatives. But the alternative is not one system that reproduces everything an organism does. It is a set of models, each rebuilding a single slice of that biology.

Continue reading below…

Like this story? Sign up for FREE Cell Biology updates:

Latest science news stories

Topic-tailored resources and events

Customized newsletter content

Subscribe

The Case for Change Is Often Made on Ethical Grounds

The case for change is often made on ethical grounds. Animal research is indeed governed by a strict framework: the principles of replacement, reduction, and refinement. These were first articulated in 1959 and have now been written directly into European Union legislation and reflected in the animal-research rules of the United States and other jurisdictions.2 That framework already constrains how animals are used, and the studies it permits have delivered real benefits. So, one has to question whether the ethical argument is the strongest ground to build from. It is a valid reason, but it is not the durable one.

The durable reason is scientific. A model made from human cells, in an architecture that resembles human tissue, can be more predictive of human biology than an animal surrogate, at least for the right questions, and it can be brought to bear early, before a compound ever reaches clinical trials. And once one accounts for how costly those trials are and how much of the total expense of drug development sits downstream of them, the argument stops being merely principled and becomes a very, very practical one. The aim, then, is to reach human-relevant answers sooner, and to reduce animal use wherever a validated alternative can carry the question.

Regulators have already begun to move on that logic. The US Food and Drug Administration (FDA) Modernization Act 2.0, which was passed at the end of 2022, removed the statutory requirement that new drugs be tested in animals and redefined the required nonclinical tests to include in vitro, in silico, and in chemico methods. It neither banned animal studies nor obliged the agency to accept alternatives, and the real movement has come since. The FDA issued a roadmap to reduce animal testing in preclinical safety studies in April 2025, starting with monoclonal antibodies, and reported meeting its first-year goals a year on. The United Kingdom followed in November 2025 with dated commitments, among them an end to animal testing for skin and eye irritation by the end of 2026. In June 2026, the European Commission adopted its own roadmap across chemical safety assessment, spanning 15 legislative domains including pharmaceuticals.

Collectively, these alternatives go by one name, New Approach Methodologies (NAMs). The field is barely standardized yet, so a name broad enough to hold all of them is itself worth something. They span computational models, organ-on-chip devices, and three-dimensional (3D) cell culture systems. The in silico methods sit largely outside the scope of this piece, which concerns the physical platforms where cells actually grow. And those are not interchangeable. Each rebuilds a different slice of human biology. So, the real question is not which platform is best, but which slice a given problem needs.

A Crowded Landscape of Platforms

Hydrogels are among the most widely used 3D platforms. Natural gels derived from collagen, fibrin, or mouse tumor-derived basement membrane extract (BME) let cells sit in a soft, water-rich environment and self-organize in three dimensions. Synthetic gels based on polyethylene glycol or similar polymers trade some of the biological cues for a defined, tunable chemistry. Their limitations track with the material. Complex animal-derived matrices such as BME vary from lot to lot, and many natural gels offer limited or poorly tunable mechanical control, whereas synthetic systems can be engineered across a wide stiffness range.3

Microfluidic and organ-on-chip systems thread cells through perfused channels that supply fluid flow and mechanical force. They can link several tissue types and reproduce dynamic conditions that static cultures cannot. The cost is complexity: low cell numbers, demanding fabrication, and a lack of standardization have slowed their wider adoption.4

Decellularized matrices strip the cells from donor tissue and aim to retain much of the native extracellular matrix (ECM), its composition, and its architecture, but the process itself can alter both.5 A related route clears the cells from plant tissue to leave a cellulose scaffold, which is inexpensive and promising, but still falls short of standardized protocols. Source-to-source variability and limited scalability remain the central obstacles.

Fibrous, or hard, scaffolds provide a solid network of fine fibers with high surface area. Because the fibers resemble the fibrillar structure of natural ECM, cells attach and behave more naturally than they do on flat plastic, and the scaffolds are sturdy enough to be cut, handled, and formatted into multiwell plates. Here the governing variable is architecture. Fiber diameter, pore size, and connectivity largely decide whether cells stay on the surface or move through the full depth of the scaffold.

Microcarriers round out the landscape. These small beads carry cells in stirred suspension and offer a large surface-area-to-volume ratio for scaling up cell numbers in bioreactors. Porous versions let cells colonize the bead interior. Their role is expansion more than tissue modeling, though even that distinction is narrowing.

Set side by side, they sort less by which is best than by which trade-off each one accepts.

Platform

Cell Arrangement

Standardization Notes

Best Suitted For

Main Limitation

Hydrogels

Embedded in 3D

Defined synthetic systems can be highly reproducible; complex natural matrices may show lot variability

Self-organization and encapsulation

Limited or poorly tunable mechanics in natural gels

Microfluidic and organ-on-chip

Patterned in perfused channels

Platform-dependent; standardization frameworks still evolving

Dynamic and multi-tissue models

Small tissue volume, complexity

Decellularized matrices

Repopulated through native structure

Source- and process-dependent

Native composition and architecture

Sourcing variability and scale

Fibrous scaffolds

Surface or depth, by pore architecture

Manufacturing- and material-dependent; can be tightly controlled with defined processing

Tissue-volume models, screening formats

No built-in perfusion

Microcarriers

On the surface or within pores

Established manufacturing; bead- and process-dependent

Scaling up cell numbers

Expansion more than modeling

Table 1. The major NAM platform categories are compared by the parameters that most influence adoption, where the choice of platform depends on which property a given application most requires.

The Whole-Organism Problem

Group the platforms by the biological scale each can represent, and a clear divide appears. Most NAMs are at their most mature for well-defined, tissue-level endpoints. A validated reconstructed-skin model can assess defined irritation endpoints, a liver construct can flag a metabolic liability, and a tumor model can screen a compound against the cells it is meant to kill. These are tissue-scale questions, and here a well-built 3D model earns its place: it answers them with human cells rather than an animal surrogate. The difficulty starts when the question turns systemic.

How a drug distributes through a body, how one organ’s metabolite acts on another, how an immune system responds as a whole—all of it depends on an integration that no isolated model reproduces. Human in vitro assays for absorption and metabolism, physiologically based pharmacokinetic modeling, and multi-organ chips are capturing more of it each year.6 But reproducing whole-body pharmacokinetics and inter-organ effects in full remains one of the hardest problems to solve without an intact organism. It is telling that the phase-out plans move slowly precisely here. The United Kingdom’s roadmap targets only a partial reduction of pharmacokinetic studies in dogs and nonhuman primates by 2030, long after skin and eye tests are gone.

The honest reading is that NAMs are furthest along where biology is local and self-contained, cosmetics and topical testing among them, and least ready where it is distributed across a whole organism. A comparison that ignored this would only flatter the technology. In that sense, knowing where a platform sits on that local-to-systemic axis is the first cut in choosing one.

Inside Fibrous Scaffolds

Because fibrous scaffolds turn on architecture, the method used to make them matters more than it first appears. Many fabrication routes exist. Three of them illustrate the trade-off well, and each leaves a different structural signature.

Electrospinning is the established workhorse. An electric field draws a polymer solution into fine fibers that collect as a mat. Conventional electrospun mats pack into dense networks whose effective pore sizes often fall below the roughly 10 to 20µm span of a typical mammalian cell, which can strongly restrict infiltration and leave cells growing as a surface layer rather than moving into the depth.7 Strategies exist to open the pores, but the baseline limitation is well documented. Solvent handling and throughput add constraints of their own.

Freeze-drying sublimes the ice from a frozen polymer solution and typically produces an interconnected porous sponge rather than a continuous nanofibrous network, though the morphology depends strongly on formulation and freezing conditions.

Another solution-spinning method is halospinning, which forms fibers from a polymer solution without an electric field. Depending on the method, it can build a more open, interconnected network that cells colonize through the full thickness. The patented halospinning process developed by Gelatex Technologies, the company of which I am the chief business development officer, is one such approach, engineered to produce an open 3D fiber architecture across a range of carrier polymers and built to be colonized through its depth rather than across its surface.

These routes expose a real tension. To welcome cells, a scaffold needs pores large enough for them to enter. To mimic ECM, it needs fibers fine enough to approach native collagen, which is organized hierarchically, from fibrils measuring tens to a few hundred nanometers across up to larger fibers. The two requirements pull in opposite directions, and the fabrication method is what has to reconcile them. When it does, cells move into the scaffold, take on lifelike shapes, form contacts in three dimensions, and lay down their own matrix, instead of flattening against a barrier.

Figure 2. Confocal microscopy of BHK-21 cells are shown after 24 hours of static culture on a halospun nanofibrous scaffold, with nuclei in blue and F-actin in red. In the orthogonal projection (left), the side views along the top and right edges capture cells at multiple depths through the scaffold. The volumetric view (right) is reconstructed from the stack of confocal optical sections, not a synthetic render, and shows the same distribution through the full depth of the material rather than a surface layer. Axes and scale are in micrometers.

Figure 3. Fluorescence microscopy of four cell lines cultured on the same halospun poly(lactic-co-glycolic) acid (PLGA) scaffold are each shown at a proliferation timepoint. One scaffold architecture supports attachment and growth across diverse cell types. Original magnification is 10×.

Supporting Organoids and Tissue Models

Organoids, which are self-organizing three-dimensional structures grown from stem, progenitor, or primary cells, sharpen the comparison between substrates, because what surrounds an organoid changes what it becomes. Many widely used protocols culture them inside drops of soft, tumor-derived matrix. That material is permissive and lets an organoid form, but it is undefined, varies from lot to lot, and offers limited, poorly tunable mechanical control. Therefore, the structures that result can differ from one experiment to the next.

A fibrous scaffold sits at the other end of the same trade-off. It gives an organoid or an engineered tissue model a defined architecture and physical support. It can be handled and formatted reproducibly, and, when its own composition and architecture are controlled, it can strip out much of the matrix-associated lot variability. What it does not give back is the permissive softness of a gel drop, the kind organoids are grown in. So, the comparison is not about which substrate is better in the abstract, but about which property a given model needs most: the self-organizing freedom of a gel or the definition and reproducibility of a scaffold.

From Bench to Screening Deck

A platform that performs in a single dish has still not proven much. Routine use is the real test, and three requirements separate a promising material from an adoptable one. The first is reproducibility at scale. High-throughput screening (HTS) depends on every well behaving like its neighbors, on compatibility with automated liquid handling, and on assay readouts that survive the move into a thicker, light-scattering 3D format, where imaging through depth often forces a choice between optical clearing and live monitoring.

The second is manufacturing. Moving from a few hand-made constructs to a reproducible commercial supply is a materials-engineering problem in its own right, and the fabrication method sets the ceiling. A process that is hard to control from batch to batch will struggle to support a qualified assay, no matter how good its biology looks at lab scale. The third is regulatory positioning. A platform whose structure and composition can be specified and controlled is easier to characterize and standardize, which can help toward qualification, though regulatory acceptance ultimately relies on demonstrating performance and biological relevance within a defined context of use. Cost then decides how widely a validated method spreads.

Where the Field Is Heading

As 3D models move from bespoke research tools toward qualified instruments for drug discovery and safety testing, the differences between platforms stop being academic. Important challenges remain across the field: vascularization, long-term functional stability, and the systemic integration discussed earlier, especially when these capabilities must be combined reproducibly in a single platform.

Closing those gaps means moving from static, single-tissue cultures toward more capable systems. Perfusion bioreactors can drive nutrients and signals through a construct instead of relying on diffusion, and maturation protocols that apply mechanical, electrical, or biochemical cues over time can coax cells toward a more functional, tissue-like state. None of this reproduces an organism, but each step narrows the distance between a dish and a body.

The pull to keep going is strong. A regulatory system is opening to alternatives, and frontiers from aging biology to human-specific disease stand to gain from models that capture what animal systems do not fully reproduce. What unites these efforts is a single aim, reaching human-relevant answers earlier, before a candidate ever enters a person. That aim is also what makes collaboration between platform types worth pursuing, since no single system reaches it alone. However, how they combine is a subject of its own. For now, the task is narrower. Match each platform’s strengths and limits to the question being asked. That match, more than any single technology, will decide how faithfully the next generation of human-relevant models reflects the biology they stand in for.

Disclosure of Conflicts of Interest

The author is employed by Gelatex Technologies, which develops the halospun nanofiber scaffolds described in the section on fibrous scaffolds.

Source link