Multiplexed and programmable regulation of gene networks with an integrated RNA and CRISPR/Cas toolkit in human cells

Abstract

RNA-based regulation and CRISPR/Cas transcription factors (CRISPR-TFs) have the potential to be integrated for the tunable modulation of gene networks. A major limitation of this methodology is that guide RNAs (gRNAs) for CRISPR-TFs can only be expressed from RNA polymerase III promoters in human cells, limiting their use for conditional gene regulation. We present new strategies that enable expression of functional gRNAs from RNA polymerase II promoters and multiplexed production of proteins and gRNAs from a single transcript in human cells. We use multiple RNA regulatory strategies, including RNA-triple-helix structures, introns, microRNAs, and ribozymes, with Cas9-based CRISPR-TFs and Cas6/Csy4-based RNA processing. Using these tools, we efficiently modulate endogenous promoters and implement tunable synthetic circuits, including multistage cascades and RNA-dependent networks that can be rewired with Csy4 to achieve complex behaviors. This toolkit can be used for programming scalable gene circuits and perturbing endogenous networks for biology, therapeutic, and synthetic biology applications.

Figure 1

The ‘triplex/Csy4’ architecture (CMVp-mK-Tr-28-g1-28) produces functional gRNAs from RNAP II promoters while maintaining expression of the harboring gene. (A) gRNA1 was flanked by two Csy4 recognition sites (‘28’), placed downstream of an mKate2 gene followed by an RNA triplex, and produced by CMVp. Csy4 generates gRNAs that can be incorporated into a transcriptionally active dCas9-VP64 (taCas9) to activate a synthetic promoter (P1) driving EYFP (P1-EYFP). (B) The presence of Csy4 enabled a 60-fold increase in EYFP levels, validating the generation of functional gRNAs. Fluorescence values were normalized to the maximum respective fluorescence between the data in this figure and Figure 2B–D to enable cross comparisons between the ‘triplex/Csy4’ and ‘intron/Csy4’ architectures. (C) Csy4 and taCas9 have opposite effects on mKate2 fluorescence generated by CMVp-mK-Tr-28-g1-28. The mKate2 expression levels were normalized to the maximum mKate2 value observed (Csy4 only) across the four conditions tested here. (D) The human RNAP II promoters CXCL1, H2A1, and UbC, and the viral CMV promoter were used to drive expression of four different gRNAs (gRNA3-6, Table S2) previously shown to activate the IL1RN promoter (Perez-Pinera et al., 2013) from the ‘triplex/Csy4’ construct. These results were compared to the RNAP III promoter U6p driving direct expression of the same gRNAs. Four different plasmids, each containing one of the indicated promoters and gRNAs 3–6, were co-transfected along with a plasmid encoding taCas9 and with or without a plasmid expressing Csy4. Relative IL1RN mRNA expression, compared to a control construct with non-specific gRNA (NS, CMVp-mK-Tr-28-g1-28), was monitored using qRTPCR. (E) The input-output transfer curve for the activation of the endogenous IL1RN loci by the ‘triplex/Csy4’ construct was determined by plotting the mKate2 levels (as a proxy for the input) versus the relative IL1RN mRNA expression levels (as the output). Tunable modulation of endogenous loci can be achieved with RNAP II promoters of different strengths, with the presence of Csy4 greatly increasing activation compared with the absence of Csy4. The IL1RN data is the same as shown in D). Data are represented as mean +/− SEM. See also Figure S1–2.

Figure 2

The ‘intron/Csy4’ architecture (CMVp-mK_EX1-[28-g1-28]_intron-mK_EX2) generates functional gRNAs from introns in RNAP-II-expressed transcripts. (A) gRNA1 is flanked by Csy4 recognition sites and encoded within an intron, leading to functional gRNA1 generation with Csy4 and activation of a downstream P1-EYFP construct. In contrast to the ‘triplex/Csy4’ construct in Figure 1, the ‘intron/Csy4’ architecture results in decreased expression of the harboring gene with increased Csy4 levels, which may be due to cleavage of pre-mRNA prior to splicing. (B–D) Three introns, a consensus intron (B), snoRNA2 intron (C), and an HSV1 intron (D), combined with Csy4, resulted in functional gRNAs as assessed by EYFP expression. Fluorescence values were normalized to the maximum fluorescence between this data and Figure 1B. Data are represented as mean +/− SEM. See also Figure S1–3.

Figure 3

Ribozyme architectures expressed from CMVp can produce active gRNAs. (A) gRNA1 was flanked with hammerhead (HH) and HDV ribozymes and encoded downstream of mKate2 with an RNA triplex (CMVp-mK-Tr-HH-g1-HDV). (B) gRNA1 was flanked with HH and HDV ribozymes and encoded downstream of mKate2 with no RNA triplex (mK-HH-g1-HDV). (C) gRNA1 was flanked with HH and HDV ribozymes (HH-g1-HDV). (D) The three architectures efficiently generated gRNA1 to activate P1- EYFP. The ‘triplex/Csy4’ construct (CMVp-mK-Tr-28-g1-28), with and without Csy4, and the RNAP III promoter U6p driving gRNA1 (U6p-g1) are shown for comparison. Data are represented as mean +/− SEM. See also Figure S4.

Figure 4

The ‘triplex/Csy4’ and ‘intron/Csy4’ architectures enable multiplexed gRNA expression from a single transcript and compact encoding of synthetic circuits with multiple outputs. (A) In the first design (Input A, ‘intron-triplex’), we encoded gRNA1 within a HSV1 intron and gRNA2 after an RNA triplex. Both gRNAs were flanked by Csy4 recognition sites. Functional gRNA expression was assessed by activation of a gRNA1- specific P1-EYFP construct and a gRNA2-specific P2-ECFP construct. (B) In the second design (Input B, ‘triplex-tandem’), we encoded both gRNA1 and gRNA2 in tandem, with intervening and flanking Csy4 recognition sites, downstream of mKate2 and an RNA triplex. Functional gRNA expression was assessed by activation of a gRNA1- specific P1-EYFP construct and a gRNA2-specific P2-ECFP construct. (C) Both multiplexed gRNA expression constructs efficiently activated EYFP and ECFP expression in the presence of Csy4, thus demonstrating the generation of multiple active gRNAs from a single transcript. Data are represented as mean +/− SEM. See also Figure S5.

Figure 5

Multiplexed gRNA expression from a single, compact transcript enables efficient activation of endogenous loci. (A) Four different gRNAs (gRNA3-6) were multiplexencoded in tandem, with intervening and flanking Csy4 recognition sites, downstream of mKate2 and an RNA triplex (CMVp-mK-Tr-(28-g-28)_3–6). (B) The multiplexed mK-Tr-(28- g-28)_3–6 construct exhibited high-level activation of IL1RN expression in the presence of Csy4 compared with the same construct in the absence of Csy4. Relative IL1RN mRNA expression was determined based on a control construct with non-specific gRNA1 (NS, CMVp-mK-Tr-28-g1-28) expressed via the ‘triplex/Csy4’ architecture. For comparison, a non-multiplexed set of plasmids containing the same gRNAs (gRNA3-6), each produced from separate, individual plasmids (CMVp-mK-Tr-28-gRNA-28) with the ‘triplex/Csy4’ architecture is shown. Data are represented as mean +/− SEM.

Figure 6

Multi-stage transcriptional cascades can be implemented with our CRISPRTF architectures. (A) A three-stage transcriptional cascade was implemented by using intronic gRNA1 (CMVp-mK_EX1-[28-g1-28]_HSV-mK_EX2) as the first stage. gRNA1 specifically targeted the P1 promoter to express gRNA2 (P1-EYFP-Tr-28-g2-28), which then activated expression of ECFP from the P2 promoter (P2-ECFP). (B) A three-stage cascade was implemented by using a ‘triplex/Csy4’ architecture to express gRNA1 (CMVpmK- Tr-28-g1-28). gRNA1 specifically targeted the P1 promoter to express gRNA2 (P1- EYFP-Tr-28-g2-28), which then activated expression of ECFP from P2 (P2-ECFP). (CD) The complete three-stage cascade from A) and B), respectively exhibited expression of all three fluorescent proteins. The removal of one of each of the three stages in the cascade resulted in the expected loss of fluorescence of the specific stage and dependent downstream stages. Data are represented as mean +/− SEM. See also Figure S6.

Figure 7

CRISPR-based transcriptional regulation can be integrated with mammalian microRNAs and RNA processing mechanisms as well as with Csy4-dependent RNA processing to implement feedback loops and multi-output circuits that can be rewired at the RNA level. (A) We created a single transcript that encoded both miRNA and CRISPR-TF regulators by expressing a miRNA from an intron within mKate2 and gRNA1 from a ‘triplex/Csy4’ architecture (CMVp-mK_Ex1-[miR]-mK_Ex2-Tr-28-g1-28). In the presence of taCas9 but in the absence of Csy4, this circuit did not activate a downstream gRNA1-specific P1-EYFP construct and did repress a downstream ECFP transcript with 8× miRNA binding sites flanked by Csy4 recognition sites (CMVp-ECFP-Tr- 28-miR_8xBS). In the presence of both taCas9 and Csy4, this circuit was rewired by activating gRNA1 production and subsequent EYFP expression, as well as by separating the ECFP transcript from the 8xmiRNA binding sites, thus ablating miRNA inhibition of ECFP expression. (B–C) Csy4 expression changes the behavior of the circuit in A) by rewiring circuit interconnections. (D) We incorporated an autoregulatory feedback loop into the network topology of the circuit described in A) by encoding 4× miRNA binding sites at the 3’ end of the input transcript (CMVp-mK_Ex1-[miR]-mK_Ex2-Tr-28-g1-28-miR_4xBS). This negative feedback suppressed mKate2 expression in the absence of Csy4. However, in the presence of Csy4, the 4× miRNA binding sites were separated from the mKate2 mRNA, thus leading to mKate2 expression. (E–F) Csy4 expression changes the behavior of the circuit in D) by rewiring circuit interconnections. In contrast to the circuit in A), mKate2 was suppressed in the absence of Csy4, but was highly expressed in the presence of Csy4 due to elimination of the miRNA-based autoregulatory negative feedback. Each of the mKate2, EYFP, and ECFP levels in B) and E) were normalized to the respective maximal fluorescence amongst all tested scenarios. The controls in column 3 and 4 in B) and E) are duplicated, since the two circuits in A) and D) were tested in the same experiment with the same controls. Data are represented as mean +/− SEM. See also Figure S7.