CRISPR Clocks: The Times They Are a-Changin’

The CRISPR revolution continues to gather speed as a new report from researchers in South Korea offers a compelling CRISPR-based technology for recording temporal analog data in living systems

In the post-CRISPR era, one could be forgiven for believing there are no biological challenges outside the reach of the emerging adaptations to this genome-modifying technology. CRISPR-based therapies are touted in the popular press as the “miracle genome editing tool” long awaited by the gene therapy community. In other fields, CRISPR technology is revolutionizing disease modeling, genetic screening, agriculture, molecular imaging, viral diagnostics, among other areas.

The rapid pace of technology development in this field makes it easy to lose track of time. Luckily, a group of researchers from Yonsei University and Seoul National University led by Hyongbum Kim has a solution for time keeping, and yes, they used CRISPR. Park et al. have creatively exploited a classical CRISPR system and adapted it to the concept of the molecular clock in the context of both in vitro and in vivo mammalian cell models. ¹ The researchers present a method in Cell in which biological events can be temporally recorded into a DNA sequence with predictable kinetics according to an exponential decay model. Their proposed synthetic “DNA Clock” system opens the door to developing living cellular devices that record the timing and duration of events such as chemical or pharmacological exposure, disease progression, tissue development, or other cellular decision making.

The concept of a molecular clock was proposed by Emile Zuckerkandl and Linus Pauling in the 1960s through their study of hemoglobin molecules across species. They observed that species-specific differences in hemoglobin structure and amino acid number linearly corresponded to evolutionary time.^2,3 This concept was eventually generalized to the molecular clock hypothesis, which surmises that DNA sequences change at a relatively constant rate over time. Based on this natural evolutionary principle, Park et al. devised a much faster synthetic molecular clock that leverages the versatile DNA edit-generating CRISPR-Cas9 analog recording system to temporally track biological events in absolute time (Fig. 1).

OPEN IN VIEWER

FIG. 1.

CRISPR-Cas9–based “DNA clocks”. A library of stgRNAs that includes a PAM is integrated into the genome of mammalian cells through lentiviral transduction. After the stgRNA expression cassette is transcribed and forms a complex with Cas9, this complex homes in on its original DNA locus; Cas9 introduces double-strand breaks in the DNA and indels are introduced into the sequence through DNA repair by nonhomologous end joining (top blue panel). As this process continues, indels accumulate within the extension and gRNA genomic and transcribed sequences (middle gray panel). In Park et al.'s DNA clock system, the mutational variants in the chromosomally encoded stgRNA loci accumulate over time. The half-life of intact nonmutated stgRNA sequences is predicted computationally and confirmed empirically to follow a model of exponential decay. This DNA clock model allows for an accurate readout of time recording, which can be estimated based on the number of intact stgRNA library sequences remaining in a biological system. PAM, protospacer adjacent motif; stgRNAs, self-targeting guide RNAs.

Reading the Time

The authors utilized a self-targeting guide RNA (stgRNA) library approach in which the conventional gRNA sequence also encodes its own target sequence through the intentional incorporation of a protospacer-adjacent motif (PAM) sequence within the gRNA. This strategy is based on previous work showing that DNA edits accumulated through self-targeting CRISPR systems can be used to create unique DNA signatures that can be resolved for cell lineage tracing or to generate mutagenesis-based metrics to correlate sequential events to a particular biological stimulus.^4–7 These and other previous approaches using base editors and other CRISPR-Cas systems can generate molecular memory in response to a biological input, although they are limited in their resolution and timescale of their temporal deconvolution.^8–11 The new method ¹ uses an array of experimental and computational approaches to effectively incorporate the dimension of absolute time as a readout in response to various biological inputs.

The central hypothesis of the “DNA clock” assumes that an intact target sequence will undergo exponential decay over time in response to indels generated by Cas9 coupled with a gRNA targeting that sequence. To test this, Park and colleagues generated a lentiviral library containing almost 24,000 unique semirandomly designed stgRNAs. They transduced this library into cells stably expressing Cas9 to establish the rate of indel generation activity. From this initial screen, they were able to identify stgRNAs with relatively high indel activity and generated a secondary sublibrary to transduce into their Cas9 knockin cell line. The authors assessed the baseline kinetic frequency of indel generation, showed a consistent correlation of indel generation rate over time across independent samples, and validated an exponential decay model of indel frequency over time.

To demonstrate the robustness and reproducibility of their time estimation model, the authors conducted a series of cross-validation studies comparing estimated elapsed time, calculated empirically from measured indel frequency, versus true elapsed time. This was accomplished using data from replicates to predict the estimated time elapsed of an excluded sample. In addition, they characterized the optimal requirements of cell number and stgRNA library size to maximize accuracy of the time estimation predictions and minimize relative absolute error. Collectively, this led to compelling evidence that the model can reasonably predict elapsed time in the stable Cas9-knockin cell line.

Although their synthetic “DNA clock” seems to work quite well in a stable Cas9-knockin HEK293 immortalized human cell line, the true test of robustness relies on the ability of this model to accurately estimate time in other systems. To this end, Park et al. conducted a series of experiments in HEK293 cells comparing three different promoters driving polyclonal lentiviral expression of Cas9 along with the focused stgRNA library. These experiments consistently underestimated elapsed time in a promoter-specific linear manner. This result is somewhat expected as the experiments were conducted in a heterogeneous cell population likely expressing various levels of Cas9 integrating at random spots in the genome compared with the relatively uniform expression of the Cas9-knockin cell line. However, the relationship between estimated elapsed time and true elapsed time remained linear, allowing the authors to define a “relative time constant” to facilitate improved accuracy in predicting true elapsed time. The mean relative absolute error between true and predicted half-life-based time estimations varied across promoters, further supporting the role of Cas9 expression level in determining clock speed. To further test the predictability of their model in other cell systems, the authors recapitulated a similar set of experiments in mouse embryonic fibroblasts and demonstrated similar results with a mean relative absolute error of ∼13–14%, consistent with their other model systems.

Following this set of rigorous systematic experiments to establish the robustness of this temporal recording method in a variety of two-dimensional cell culture systems, the authors tested whether it would perform similarly in different cellular microenvironments and cell types in vivo. Effective prediction of elapsed time based on indel frequency generation was readily achieved with the Cas9-knockin cells cultured on three-dimensional scaffolds in vitro or implanted subcutaneously in mice, suggesting the system is resilient to changes in culture environment.

To challenge their DNA clock system in a less controlled environment, Park et al. delivered their lentiviral sublibrary into Cas9-knockin mice through intradermal injection and analyzed skin tissue at the site of administration, revealing a linear correlation between predicted and elapsed time within a complex tissue containing a variety of resident cells types. The mean relative absolute errors had a much broader range in this context, though as with the other experimental systems, the error level seemed to stabilize after 10 days.

Fast Forward

The synthetic “DNA clock” CRISPR-Cas9 system presented by Park et al. is an elegant combination of theory based in evolutionary biology and a beautifully simple CRISPR-based technology. Nevertheless, the potential impact of this system lies in its utility to resolve yet unanswered biological questions. The authors conclude their article with a series of proof-of-principle applications showing how this system can be coupled with different methods to drive Cas9 expression. This includes chemical induction of Cas9 expression and heat- and inflammation-responsive promoters driving expression of Cre recombinase that activates expression of the stgRNA cassette and Cas9, delivered with a Sleeping Beauty transposon system. In a particularly compelling example, they extend the inflammation-responsive approach into mice with implanted cells that track the duration since exposure to an inflammatory stimulus. Across this variety of experimental systems, the authors show that their method can reliably use CRISPR-mediated targeted indel generation to estimate elapsed time in response to an initial stimulus in vitro and in vivo.

As the diversity of researchers working with CRISPR-Cas9 systems in their research continues to expand, this DNA clock could be envisioned to impact several areas. Using this system along with tissue-specific promoters could generate previously unrecordable data in the context of pathophysiology studies, lineage tracing studies of unprecedented temporal resolution to inform developmental biology, as well as temporal tracking of metastasis in cancer research.

Importantly, there are several caveats on how this system might effectively be applied across a wide range of cell types and in vivo models. The stgRNA libraries may need to be reoptimized for more robust time keeping in vivo with other cell types, delivery vehicles, and expression cassettes. Some stgRNA sequences have measured half lives up to 2–3 years, although the longest time point reported in this study is 60 days. It will be interesting to observe the stability of this system over longer periods. Nevertheless, Park et al. present a novel compelling CRISPR-based technology to address unresolved challenges in recording temporal analog data in living systems, thus continuing to expand the scope of the CRISPR revolution.