This article was published online on September 2 in the journal Nature cell biology by researchers at the University of Pompeii Fabra in Spain and ICREA Systems Bioengineering.

基于干细胞的胚胎模型标准化标准（全文）

Pluripotent stem cells are being used to generate early embryogenesis models that show great promise in discovery and transformation research. In order for these models to be of practical value, we need to strictly consider their efficiency and loyalty to natural embryos. Here, we propose guidelines to improve the standards for research on stem cell-based embryonic models.

The word "model" has different contextual meanings in science. In physics and engineering, a model is usually a simplification of reality, usually a mathematical representation, used to explore the interactions between various parts of a system to understand overall behavior. In developmental biology,"model" refers to the use of one organism to help clarify general principles of development in multiple organisms. Due to the conservation of many processes among species, the use of model organisms for research has certain legitimacy. In mammalian research, laboratory mice are traditional model organisms. However, due to differences in morphogenetic processes, structure and timing among different species, it is unclear to what extent the knowledge gained from mice can be extended to other mammals, especially humans.

Research on human embryos has a history of more than 100 years, but is mainly descriptive rather than institutional research. Interest in studying early human development has been re-stimulated in recent years as donations of excess human embryos resulting from in vitro fertilization have increased. This has led to some new insights into the development of human embryos to gastrulation (approximately day 14). However, after this stage, in vitro experiments on human embryos still present technical difficulties and are prohibited in many countries/regions. Therefore, conducting institutional research on human embryos remains a huge challenge. Although some countries, such as China and Israel, have fewer restrictions on human embryo research, the current regulatory framework for human embryo research limits progress in understanding how human embryos occur in healthy and disease states.

In the past two decades, pluripotent stem cells (such as embryonic stem cells and induced pluripotent stem cells) have emerged as tools for studying mammalian development. Adherent cell cultures and three-dimensional (3D) embryoid bodies produced by pluripotent stem cell cultures are capable of displaying characteristics of genetic and epigenetic programs, such as responses to cytokines, that mimic the assignment, determination and differentiation of cell lineages during natural embryonic development. These cultures are often referred to as "developmental models" and are useful in producing relevant cell types for scientific research and therapeutic purposes, such as drug discovery and cell therapy. However, these models lack a complete representation, proportion and organization of cell types during natural three-dimensional development of embryos.

Over the past decade, groundbreaking research has revealed the ability of pluripotent stem cell aggregates to self-organize and form patterned structures (called organoids) that mimic the structure of the origin of an organ in vitro. Since then, numerous protocols using pluripotent stem cells and precise signaling schemes have been shown to reproduce the cell lineage trajectories observed in "developmental models" and further incorporate morphogenesis and tissue modeling processes unique to natural mammalian organogenesis. The self-organizing capabilities of pluripotent stem cells have recently been further utilized to form structures similar to early mammalian embryos, including human embryos, which has led to a simplified, bottom-up approach to simulate human embryogenesis. These mammalian embryo models do not require the generation of precise replicas of natural embryos to be useful in research. However, they should be close enough to their in vivo counterparts, including, for example, the correct constituent cell types, and show the structural organization of a natural embryo, while being easy for experimental manipulation to provide useful new knowledge about development. Based on this paradox, embryonic models may be particularly useful if they form only certain isolated tissue groups, revealing autonomous processes that may be concealed in the complex integrated environment of natural embryos. Embryonic models may also follow a different path from the classic development path because their cellular functional attributes may be less restrictive than in embryos; this may provide valuable information. Therefore, the main goal of embryo modeling is not necessarily to generate complete replicas of embryos or their constituent structures, but rather to illuminate specific aspects of development by leveraging their scalability, accessibility, modularity and experimentability.

A particularly useful class of embryonic models for the study of human development is those derived from human pluripotent stem cells. However, due to the ethical sensitivities posed by their similarities to natural embryos, discussions have been sparked about whether scientific inquiry can justify the use of these human embryo models in research. We believe that their relevance should be judged by weighing their potential benefits against models with less ethical burden, such as developmental models or organoids. In addition, their usefulness depends on their ability to effectively reproduce developmental events such as tissue patterning and morphogenesis. Where the results and positive impacts of scientific understanding transcend ethical and legal issues, modeling a complete embryo may be justified, provided regulatory guidelines are followed. In this case, the generation and use of embryonic models should be monitored to ensure that these models avoid features that would be excluded from natural embryonic research for ethical reasons. However, we speculate that if an embryo model becomes indistinguishable from natural embryos at some point, it should be subject to the same rules and legislative windows that apply to embryo research.

Criteria for defining stem cell-based embryonic models

Embryo models may show different levels of production efficiency, cell composition and structural organization, but have low loyalty to real embryos. This situation is problematic because it can lead to erroneous representations of the model and thus risk misleading scientific and medical findings. For example, over-reliance on insufficiently characterized models can lead to erroneous discoveries in embryology and misleading use in disease modeling or drug testing. Therefore, we believe it is necessary to propose some basic attributes to characterize embryonic models.

A key requirement for a model is that its characteristics should reflect the characteristics of the entire or part of the embryonic structure being simulated, and this should be clearly stated in the report of the research results.

In order to make experiments and discoveries repeatable, the report should answer the following questions: What is the pluripotent state of the initial pluripotent stem cell population? To what extent do the constituent cells of the embryonic model structure reflect the cellular state and spatial organization of the embryo? How often is the successful modeling of embryonic goals? If the final model structure deviates from the goal, it is important to acknowledge these differences and determine the specific points at which they deviate from the true state of embryonic development. In addition, the development path of the embryo model should be recorded because there are some important issues to consider, such as: the extent to which the cells undergo the same sequence and rate of cellular state and morphological changes as the embryo, which stages of development have cells and tissues gone through over time, the degree of synchrony in the model, and whether the model skips certain events, accelerates or slows down the development process compared to the progress of the embryo.

The efficiency, repeatability and robustness of the model are key requirements for using embryonic models as useful experimental tools to clarify embryogenesis or for transformation research. Therefore, reporting of these variables should be carried out through robust statistical measures to quantify relevant characteristics, and publications must be accompanied by reports of such information.

Methods to characterize embryonic models

The loyalty of an embryo model can be characterized by benchmarking its cell composition and transcription status with a reference data set, usually compared with the cell-type transcriptome of a natural embryo or its body parts and organs at a specific stage of development. This analysis, supported by improved visualization methods, is key to understanding the similarities and differences between embryonic models and natural embryos. When projecting the model's transcriptome data onto the reference transcriptome dataset, those cells that are not typical of the target tissue, residual pluripotent stem cells or progenitor cells should be included to ensure that projection-based tag transfers do not overfit.

Gene expression profiles are useful proxies for characterizing models of preimplantation development stages, because during this period the number of lineages is small and the organization is relatively simple. However, as development progresses, the value of gene expression descriptors gradually decreases. For example, during gastrulation, the range of cell types expands significantly, and the catalog of cell types alone is no longer an important criterion, and spatial arrangement and relative proportions become more critical. For example, a two-dimensional culture model of gastrula produces an array of cell types similar to those present in gastrula. Therefore, its topological organization and functionality should become the main criteria for comparing models and embryos. Future efforts should aim to improve the granularity of the reference dataset by integrating additional layers of stage-specific multi-omic information, such as spatial organization of cell types, epigenomes, proteome, and metabolome characteristics.

In general, accurate and quantitative descriptions of specialization, differentiation, pattern formation, and morphogenesis events in space and time are needed to assess the loyalty and repeatability of the model, as well as its ability to simulate embryogenesis. This is critical to comparing different models, defining the range of scientific and medical problems that can be solved with them, and selecting the most useful model to solve specific problems. We realized that such detailed characterization cannot be done in a single study, but rather requires the joint effort of the entire research community to comprehensively characterize the models at stake.

Develop terminology

Model naming is an important descriptor that can influence the perception of research work and the attitude of funding agencies, especially in countries such as the United States, where funding for research on embryon-like structures, such as structures composed of trigerm layers or their derivatives, is subject to close federal review and is currently decided on a case-by-case basis by the National Institutes of Health. The "universal" name for an embryonic model should reflect the organization of the system and the identity of the objects it is intended to simulate. We suggest that, first of all, the model should use universally understood terms. Secondly, it should reflect the stages and organization being simulated, and ideally should also suggest (if necessary) the imperfect nature of the model. This is often the original intention of using the suffix "-oid" to indicate that something is similar but not equivalent.

Accurate terminology will help improve clarity of the work being carried out, both in the field and for the public's correct understanding. For example, one of the problems that arises in the current situation is that different names are given to models that reproduce aspects of tissue patterning related to the proto-axis in the paraxial mesoderm: somitoids, segmentoids, axioloids, and trunk-like structures. Although each model has specific characteristics, and some of the segments without axial organization may be considered organoids, most attempt to represent the same segment formation process and associated embryonic structures. However, multiple names for similar or related objects can lead to confusion, so uniform terminology should be identified to promote dialogue among scientists and, more importantly, to promote media and society's understanding of the field. We recommend that this task be accomplished by consensus among the researchers developing the model.

As for an overarching term, several names have been proposed, such as stembryos, embryoids, and pseudoembryos, each with its own values and shortcomings. All of these names have the advantage of suggesting that the embryo model is an imperfect replica of a natural embryo, but may also be misinterpreted as being close to the embryo. We acknowledge that naming issues require consensus through discussions within the scientific community that are beyond the scope of this review. However, we note that the term "stem cell-based embryonic models" is currently gaining recognition, and we recommend following this term. We also suggest that, where appropriate, specific features of embryogenesis simulated in the model (such as gastrulation, cardiogenesis, axial elongation, and patterning) can be added as "suffixes" to clearly define the attributes of a particular model; for example, stem-cell-based gastrulation models, stem-cell-based somite formation models.

Conclusion

Our goal is to emphasize the importance of clearly and consistently describing the efficiency and accuracy of embryonic models to define their practicality, and to reach consensus on terminology to improve communication. To this end, we put forward some suggestions on experimental standards and reports. Appropriate characterization of embryo models will facilitate review by funding agencies on the value of research and review by regulators, allowing progress in the field based on public trust.

Attachment: Attributes of embryo models

Generation of embryonic models

1. Starting materials:

a. The pluripotency status and genomic integrity of the stem cell line should be reported.

b. Ideally, multiple different cell lines should be used and compared.

c. Report the initial number of cells used for embryo modeling and other cell types, if applicable.

2. Agreement:

a. The method of assembling the starting cell population to initiate model generation should be described in detail.

b. Step-by-step in vitro culture conditions should be described in detail to achieve modeling endpoints.

3. Quantitative measures of the efficiency of model generation and repeatability of final characteristics should be reported in a statistically accurate manner.

Characterization of embryonic models

Foreword: Define the goals of modeling, such as blastocyst, double-layer embryo disc, 14-day embryo, gastrula, body axis pattern, etc.

1. Identify the cell composition and spatial organization in the model.

2. Assessment results: Morphological characteristics of cell structure.

3. In the context of the findings of points 1 and 2 above, determine the level of loyalty for specific target modeling, using the benchmark standards described below.

4. Quantitatively measure intra-experimental and inter-experimental variability of modeling results.

5. Clarify the limitations of the model.

基准标准

1. Cell composition and cell state, determined from transcriptome data and, if applicable, additional methods (e.g., proteome, metabolome).

2. Spatial organization of cell types in the model structure, and (such as related) substructures (e.g., somites, neural tubes).

3. The shape of the complete structure and, if relevant, its components (e.g., individual organ primordia).

4. The spatio-temporal sequence of morphogenetic events.

5. Matching goals based on points 1-4 with development stages.

Further reporting standards

1. Similarities and differences between the embryo model and the natural embryo target structure should be evaluated and reported. For classifier-driven annotations, confidence scores should be reported.

2. Limitations of the model, such as non-target cell types and morphological variations, should be clarified.

3. For embryo models that require scientific and regulatory supervision, provide ethical and regulatory governance statements for the generation and use of embryo models for research.