A CONTROLLABLE HUMAN SPINAL CORD MODEL WITH FULL DORSOVENTRAL PATTERNING IDENTIFIES DORSALIZING ROLE OF RETINOIC ACID

Introduction: Understanding embryonic development has vast scientific and translational implications. During development, cells must differentiate into different cell types according to their positions. Animal studies have discovered that positional information is conveyed through concentration gradients of signaling molecules called morphogens. A canonical model system for morphogen gradient-driven patterning is the dorsoventral (DV) patterning of the spinal cord (SC). The spinal cord is exposed to antiparallel gradients of dorsalizing and ventralizing morphogens, including BMP4 and SHH, respectively. Through interpreting the gradients, the spinal cord is patterned into 13 transcriptionally distinct domains along its DV axis.
Advancements in stem cell technologies have enabled the development of human pluripotent stem cell (hPSC)-based developmental models promising for elucidating the mechanisms underlying human development and associated disorders. Various spinal cord DV patterning models have been developed, but they lacked the antiparallel morphogen gradients, exhibiting partial patterning and suffering from low efficiency and controllability. Herein, we report the development of a microfluidic, hPSC-based system to recapitulate SC DV patterning by recapitulating the antiparallel signaling gradients. The system exhibited controllable patterning, and all 13 transcriptionally distinct domains were present.

Materials and

Methods: Generation of spinal cord-like structures (SCLS)
First, hPSCs were patterned into rectangular colonies inside a microfluidic channel using microcontact printing. Then, the channel was filled with hydrogel, and neural induction medium was supplied. Upon hydrogel overlay and neural induction, the monolayer colonies turned into 3D tubular structures with spinal cord-like identity, forming spinal cord-like structures (SCLS). To achieve DV patterning, proper signaling molecules were added to either opening of the channel. The molecules diffused into the hydrogel-filled channel, forming signaling gradients. For dorsal morphogens, recombinant BMP4 (R&D System) was used. For ventral signals, either the BMP pathway antagonist LDN193189 (STEMCELL Technologies) or the Hedgehog pathway agonist SAG (STEMCELL Technologies) was used.

Identification of 13 DV domains
To identify 13 distinct cell types along the DV axis, we performed single-cell RNA sequencing on our SCLSs using the 10X Genomics Chromium system and Illumina NovaSeq-6000. Downstream analysis was done using the R package Seurat (v5.0.1, https://satijalab.org/seurat/). Cells were first partitioned into broad lineages, including neurons, neural crest cells, and neural progenitor cells. Neural progenitor cells were examined for the expression of known marker genes and were allocated a DV domain identity accordingly.

Results, Conclusions, and Discussions: The application of signaling gradients resulted in robust and controllable patterning. When BMP4 was added to the dorsal well, and the BMP pathway antagonist LDN193189 was added to the ventral well, over 95% of SCLS showed patterned expression of the dorsal marker PAX3 and intermediate marker PAX6 (Fig.1a). Also, all of the samples contained SOX10+ neural crest-like cells delaminating from the dorsal pole. The patterning was controllable as the dorsal domains expanded according to an increase in BMP4 concentration. Similar results were observed when basal medium was added to the dorsal well and the Hedgehog pathway agonist SAG was added to the ventral well (Fig.1a, b). To achieve full DV patterning, we added BMP4 to the dorsal well and SAG to the ventral well. The antiparallel signaling gradients resulted in fully DV patterned SCLSs as they contained all 13 DV domains, identified by single-cell RNA sequencing (Fig.2b, c). We further confirmed that our SCLSs recapitulated human-specific transcriptional features of the SC (Fig.2d).
Capitalizing on the robustness of our SCLS DV patterning, we used our model to study the role of retinoic acid (RA) in SC DV patterning, which remains elusive since in vivo studies have shown contradictory roles of RA in regulating both dorsal and ventral transcription factors. Data obtained from our model show that RA has an overall dorsalizing effect, inhibiting the expression of ventral markers (Fig.3c, d) and expanding the dorsal marker PAX3 (Fig.3e, f).
Together, we have demonstrated that the recapitulation of signaling gradients can greatly improve the robustness and controllability of a hPSC-based SC model, which further allowed us to investigate the role of RA in SC DV patterning. Signal gradient-mediated patterning is a prevalent mechanism in embryo development. Thus, our embryo modeling technology will be useful for the study of other developmental systems.

Acknowledgements (Optional):