The expressions “biological replicate” and “technical replicate” are ubiquitous in biology, but they are surprisingly recent. At first glance, the adjectives suggest that one can understand these notions in terms of a distinction between biological variation and technological variations. This however becomes problematic as soon as biological systems are used as instruments.

Consider the example of the teratoma assay, the gold standard for assessing the differentiation potential of alleged pluripotent stem cells. In this test, the cells are injected in a mouse to see if they will grow a teratoma reconstituting the three germ layers: if they do, they are considered pluripotent. Growing and differentiating are inherently biological processes, and yet if we performed two teratoma assays using cells from the same dish, we would generally call these experiments 'technical replicates'. The reason is that the mice are here a sort of measuring device, used to assess a property of the injected cells.

If one looks up the terms in the NIH's biomedical ontology, the examples for biological and technical replicates are, respectively, “a patient in a given arm of a trial” and “Aliquots of a tissue subjected to parallel assays”. Likewise, in textbooks the distinction will hinge on whether the experiments are performed twice on the same sample, or on two different samples. 'Sameness', here, is accomplished by random sampling: by sampling (randomly and with a reasonable size) twice from a single population of cells (a dish, or the patient's blood, etc), one obtains two populations deemed identical (an assumption which may be problematic for solid tissues). This prompts the question of whether single-cell experiments performed on cells of the same population constitute biological or technical replicates, but for the moment I will leave this question aside.

Consider another example: a prospective stem cell therapy, in which we inject stem cells (coming from the same population) into the brain of two patients affected by a neurological disease, in order to see whether the intervention alleviates the disease. Clearly, we would not consider the two experiments as technical replicates, despite the fact that both patients were injected random samples of the same population of cells. Instead, we would consider that it is the two patients that represent two samples, and therefore that the two experiments are biological replicates. But why, in the case of the teratoma assay, did we not consider the two mice as two samples? Or why, in the case of stem cell therapy, did we not consider the patients are measuring devices, meant to assess the regenerative properties of the injected cells?

The core difference is that in the teratoma assay, a robust link was already established between the pluripotency of the cells and their ability to grow teratoma under certain experimental conditions. As a consequence, the process of engraftment, growth and teratoma formation can be black-boxed, and the mouse can be used as a mere instrument for the investigation of the injected cells. In other words, the mouse and the assay can be considered as a technical object, while the injected cells are part of theepistemic thing – that which we are currently trying to understand (1).

In contrast, in the case of the stem cell treatment, it is not the stem cells that are the object of investigation, but the intervention performed with them. The intervention and the processes it involves cannot be black-boxed and used as a technical object, because they have not yet been established.

If this analysis is correct, then what in these circumstances are considered biological replicates could, in other circumstances, be considered technical replicates.

Suppose, for instance, that a given stem cell therapy was shown to robustly improve a given pathological condition. However, in order to make the therapy more accessible and reduce risks of rejection, we are investigating whether induced pluripotent stem cells (iPSC) are as efficient as embryonic stem cells for this treatment: we compare the outcome of the treatment using different population of cells, reprogrammed through different means. In these circumstances, the treatment itself effectively becomes a measurement assay performed on the different kinds of cells. Therefore, two patients injected with the very same population of cells should be considered as technical replicates. Biological replicates would instead be two patients respectively injected with two populations of cells which were reprogrammed using the same method.

Instead of the distinction between biological and technical replicates, it might be more precise to speak, following Rheinberger's (1) distinction between “epistemic things” and “technical objects”, of a distinction between epistemic and technical replicates.

References

1. Rheinberger, H.-J. (1997). Toward a History of Epistemic Things: Synthesizing Proteins in the Test Tube. (T. Lenoir Gumbrecht H U, Ed.). Stanford University Press.