The remarkable development of a single cell, the zygote, into the full organism occurs through a complex series of division and differentiation events that resemble a tree, with the zygote at the base branching through lineages that end in the terminal cell types at the top. Characterizing this tree of development has long been a subject of interest, and the combination of modern genome engineering and sequencing technologies promises a powerful strategy in its service: in vivo barcoding. For in vivo barcoding, heritable random mutations are induced to accumulate during development and sequenced post hoc to reconstruct the lineage tree. Demonstrations thus far have largely focused on lower vertebrates and have used a barcoding element with a constrained window of activity for clonal tracing of individual cells or cell types. Implementation in mammalian model systems, such as the mouse, incurs unique challenges that require major enhancements.

To address the complexity of mammalian development, we reasoned that multiple independent in vivo barcoding elements could be deployed in parallel to exponentially expand their recording power. Independence requires both an absence of cross-talk between the elements and an absence of interference between their mutation outcomes. A system with the potential to deliver on these requirements is homing CRISPR, a modified version of canonical CRISPR wherein the homing guide RNA (hgRNA) combines with CRISPR-Cas9 nuclease for repeated targeting of its own locus, leading to diverse mutational outcomes. Therefore, in mouse embryonic stem cells, we scattered multiple hgRNA loci with distinct spacers in the genome to serve as barcoding elements. With this arrangement, each hgRNA acts independently as a result of its unique spacer sequence, and undesirable deletion events between multiple adjacent cut sites are less likely. Using these cells, we generated a chimeric mouse with 60 hgRNAs as the founder of the MARC1 (Mouse for Actively Recording Cells 1) line that enables barcoding and recording of cell lineages.

In the absence of Cas9, hgRNAs are stable and dormant; to initiate barcoding, we crossed MARC1 mice with Cas9 knock-in mice. In the resulting offspring, hgRNAs were activated, creating diverse mutations such that an estimated 10²³ distinct barcode combinations can be generated with only 10 hgRNAs. Furthermore, hgRNAs showed a range of activity profiles, with some mutating soon after conception while others exhibited lower activity through most of the gestation period. This range resulted in sustained barcoding throughout gestation and recording of developmental lineages: Each cell inherits a set of unique mutations that are passed on to its daughter cells, where further unique mutations can be added. Consequently, at any stage in such developmentally barcoded mice, closely related cells have a more similar mutation profile, or barcode, than the more distant ones. These recordings remain embedded in the genomes of the cells and can be extracted by sequencing.

We used these recordings to carry out bottom-up reconstruction of the mouse lineage tree, starting with the first branches that emerged after the zygote, and observed robust reconstruction of the correct tree. We also investigated axis development in the brain by sequencing barcodes from the left and right side of the forebrain, midbrain, and hindbrain regions. We found that barcodes from the left and right sides of the same region were more closely related than those from different regions; this result suggests that in the precursor of the brain, commitment to …