key: cord-0730927-zf13h0or authors: Liu, Yang; Pinto, Filipe; Wan, Xinyi; Peng, Shuguang; Li, Mengxi; Xie, Zhen; French, Christopher E.; Wang, Baojun title: Reprogrammed tracrRNAs enable repurposing RNAs as crRNAs and detecting RNAs date: 2021-05-28 journal: bioRxiv DOI: 10.1101/2021.05.24.445356 sha: 02627cbca46b3f2ee0199dff5f2d4602117f5a6f doc_id: 730927 cord_uid: zf13h0or In type II CRISPR systems, the guide RNA (gRNA) consists of a CRISPR RNA (crRNA) and a hybridized trans-acting CRISPR RNA (tracrRNA) which interacts directly with Cas9 and is essential to its guided DNA targeting function. Though tracrRNAs are diverse in sequences and structures across type II CRISPR systems, the programmability of crRNA-tracrRNA hybridization for particular Cas9 has not been studied adequately. Here, we revealed the high programmability of crRNA-tracrRNA hybridization for Streptococcus pyogenes Cas9. By reprogramming the crRNA-tracrRNA hybridized sequence, reprogrammed tracrRNAs can repurpose various RNAs as crRNAs to trigger CRISPR function. We showed that the engineered crRNA-tracrRNA pairs enable design of orthogonal cellular computing devices and hijacking of endogenous RNAs as crRNAs. We next designed novel RNA sensors that can monitor the transcriptional activity of specific genes on the host genome and detect SARS-CoV-2 RNA in vitro. The engineering potential of crRNA-tracrRNA interaction has therefore redefined the capabilities of CRISPR/Cas9 system. The type II clustered regularly interspaced short palindromic repeats (CRISPR) system employs a small non-coding RNA known as trans-activating CRISPR RNA (tracrRNA) to form the mature dual-RNA structure of guide RNA (gRNA) (1-5). By complementary pairing with a precursor CRISPR RNA (pre-crRNA), the tracrRNA mediates RNase III-dependent RNA processing to generate mature crRNA and Cas9 ribonucleoprotein (RNP) complex (3) (4) (5) (6) (7) (8) . The pre-crRNA consists of continuous alternating spacers and repeat sequences (9) . The unique spacers have homology to different foreign DNA sequences and guide the Cas9 RNP to recognize the target DNA by RNA-DNA pairing (8, 10) , while the repeats of pre-crRNA can form an RNA duplex with the tracrRNA by hybridizing to an anti-repeat region near its 5'-end (6) . Consequently, the crRNA:tracrRNA RNA duplex will be cleaved by RNA duplex-specific RNase III in the presence of Cas9, leading to fragmented pre-crRNA with one spacer for each mature crRNA (4, (6) (7) (8) . The tracrRNA plays a vital role in the maturation of gRNA and in the CRISPR function, and has molecular interactions with Cas9 protein (4, 11, 12) . In many applications and studies regarding CRISPR, the crRNA and tracrRNA can be artificially fused into a single gRNA molecule called sgRNA (13) . The 3'-end truncated crRNA and 5'-end truncated tracrRNA can be linked and still support the function of CRISPR/Cas9. The crystal structure of the Cas9 complex reveals that the sgRNA scaffold has strong interactions with Cas9 and plays an essential role in CRISPR function (5) . Although the function of tracrRNA is conserved in the CRISPR systems of various species, its sequence and localization within the CRISPR-Cas locus are highly diverse (8, 14) . It is also known that there is an orthogonal relationship between tracrRNAs and the corresponding Cas9 proteins from distant species (8, 13) . The relationship between the Cas9 and dual-RNA raises an intriguing question of whether the complementary pairing between crRNA and tracrRNA is programmable for particular Cas9 proteins. Several previous studies support the potential programmability of the crRNA-tracrRNA pairing. First, the Cas9 protein of specific bacteria can form a functional complex with tracrRNAs from closely related species (8, 13) . In these studies, the available heterogeneous tracrRNAs have similar secondary structures but different sequences, suggesting that Cas9 proteins may be more sensitive to the secondary structure of the gRNAs than to their specific sequences. Second, the tetraloop stem of the sgRNA, equivalent to the crRNA-tracrRNA pairing fragment, is programmable for the Streptococcus pyogenes Cas9 (SpCas9) (14, 15) . Assuming that the difference between sgRNA and dual-RNA will not change the Cas9-gRNA interaction model, the sequence of the crRNA-tracrRNA matching region should also be reprogrammable. Finally, previous studies reveal that the CRISPR/Cas9 system can tolerate single base-pair substitutions in the crRNA-tracRNA pairing region (16) , suggesting multiple base-pair substitutions may also be feasible. Nonetheless, only recently has the programmability of the crRNA-tracrRNA hybridization region been explored to detect RNA molecules (17) . 3 Here, we systematically explore the programmability of crRNA-tracrRNA pairing of the CRISPR/SpCas9 system, built on our initial intention of developing programmable AND logic devices. We validate the programmability of crRNA-tracrRNA and reveal a new set of principles to guide the design of CRISPR systems with reprogrammed dual-RNA. The high programmability of crRNA-tracrRNA pairing also brings new perspectives and potential for the engineering and application of the CRISPR/Cas9 tool. The orthogonal AND logic gates are developed based on this mechanism. Further, by reprogramming the crRNA-tracrRNA pairing, SpCas9 can specifically repurpose various RNAs as crRNAs for triggering CRISPR function. We develop an mRNA sensor able to hijack of endogenous RNA molecules as crRNAs and show that the inherent characteristics of mRNA and tracrRNA structures can strongly affect the mRNA-mediated CRISPR function. Notably, we successfully monitor the transcription level of endogenous genes in E. coli and connect the bacterial genetic network to an artificial gene circuit, which works as a whole-cell arsenic biosensor. A new type of RNA sensor is demonstrated by detecting SARS-CoV-2 target This study has thus redefined the application capabilities of Cas9 proteins and the sources of crRNAs, and provides new scope for further studying the type II CRISPR systems. To study crRNA-tracrRNA pairing, we designed a crRNA-tracrRNA mediated CRISPR activation (CRISPRa) device in bacteria modified from a previously reported eukaryote-like CRISPRa system (18, 19) . A crRNA-tracrRNA mediated CRISPR activation (CRISPRa) device requires splitting the sgRNA into crRNA and tracrRNA. However, in the original design of our CRISPRa system, one of the two RNA aptamers occupies the tetraloop of sgRNA, equivalent to the structure where crRNA and tracrRNA are connected. Therefore, splitting the sgRNA would destroy this structure and the function of our CRISPRa device (Figure 1a) . To overcome this issue, we redesigned the sgRNA by moving the RNA aptamers to the 3'-end of the sgRNA. As a result, we proved that the sgRNA with tail-fused aptamers is functional for the eukaryote-like CRISPRa system. Further, the 3aptamers design provides higher efficiency than the 2-aptamers sgRNA design (Supplementary Next, we split the sgRNA with 3'-end fused aptamers to tracrRNA and crRNA at the tetraloop position (Figure 1a) . Apart from the nine essential RNA base pairs for the functional dCas9 RNP complex (13) , additional base pairs with the same sequence as wild type (WT) crRNA and tracrRNA from Streptococcus pyogenes were designed to mimic the natural crRNA-tracrRNA pairing. The expression of dCas9, crRNA and tracrRNA were controlled by different inducible promoters (Figure 1b ). An experiment permuting three induction conditions indicated that only when crRNA, tracrRNA and dCas9 were all induced did the CRISPRa system give the highest output (Figure 1c) . When we removed the crRNA generator or tracRNA generator circuit, no activation of the CRISPRa system was observed. The activation efficiency of crRNA-tracrRNA mediated CRISPRa and sgRNA mediated CRISPRa are in the same order of magnitude ( Figure 1d ). In the above tests, we confirmed that the dual-RNA mediated CRISPR system could also be employed for eukaryote-like CRISPRa in bacteria. Therefore, we set this crRNA-tracrRNA mediated CRISPRa device as a platform for the subsequent study of the programmability of crRNA-tracrRNA pairing in the CRISPR/Cas9 system. We first reprogrammed the base pairs of the core crRNA-tracrRNA hybridizing region. The core region is the minimal matching region of tracrRNA and crRNA for CRISPR function including 12 nt RNA at the repeat region of crRNA and the corresponding tracrRNA region (13) . We introduced a stretch of mutations into the crRNA and made complementary mutations in tracrRNA for testing (Figure 1e) . Specifically, the core matching region includes nine RNA base pairs and two wobble base pairs (G-U). Two bases, GA, in crRNA and four bases, AAGU, in tracrRNA form a bulge structure, essential for CRISPR/Cas9 function (14) (Figure 1e) . Previous studies have shown that the deletion or the replacement of the nucleotides in the bulge can abolish the sgRNA function (14, 15) . Assuming that the same rule applies to crRNA-tracrRNA pairs, we reprogrammed the nine base pairs except the nucleotides of the bulge and wobble base pairs in the first step. The library of the above reprogrammed crRNA-tracrRNA pairs was tested using our CRISPRa system. The results show that all the reprogrammed crRNA-tracrRNA pairs enabled efficient CRISPRa output. Further, all the outputs were in the same order of magnitude as that of the wildtype (WT) crRNA-tracrRNA hybridizing region (Supplementary Figure 1c) . We then explored the orthogonality of mutated crRNA-tracrRNA pairs by recombining the mutated crRNAs and tracrRNAs. Interestingly, the bulge structure is like a dividing line. The upstream region near the bulge is less able to tolerate mismatches than the immediately downstream (Figure 1f ). For SpdCas9, more than two adjacent mismatches in the upstream region of the bulge inhibit the CRISPRa function. By contrast, in the downstream region, even four mismatches could be tolerated for CRISPRa function. Surprisingly, in some cases, the mismatches in the downstream region enable higher CRISPRa output than that of the WT crRNA-tracrRNA pair (Figure 1f) . Regarding the original purpose of this work, the dynamic behavior of the dual-RNA mediated CRISPRa system shows the Boolean logic profile of AND gate, providing an output only when both the inputs are present (20) . Although AND-gate circuits have various applications in synthetic biology (21) , building complex artificial genetic networks requires a large number of highly orthogonal AND gates. CRISPRa based on the programmability of crRNA-tracrRNA pairing can efficiently address this problem. In this case, the presence or absence states of the crRNA or tracrRNA represent states 1 and 0 of the inputs, respectively. The maximum and minimal expression levels of the sfGFP represent the states 1 and 0 of the output (Figure 1g) . Our experiments have exhibited a typical AND-gate function of this system in E. coli (Figure 1d) . We then randomly generated four sequences for the crRNA-tracrRNA matching region and tested the orthogonality between these AND gates. As a result, we proved that orthogonality between the AND-gate circuits could be achieved based on the principle of complementary base pairing ( Figure 1g ). As inducible systems are more desired to control logic gate devices (18), we employed the induced or non-induced states of the crRNA or tracrRNA to represent states 1 and 0, respectively. For the inducible system, this AND gate became less efficient due to its high sensitivity to tracrRNA leakiness from the inducible promoter (Figure 1c) . To overcome this issue, we truncated the crRNA-tracrRNA pairing region to 14 bp to weaken their pairing affinity. As a result, this optimization improves the AND gates with good symmetry in response to the two inputs ( Figure 1h ). In addition, inspired by our previous experience with dxCas9 3.7, and in how it could optimize the sensitivity to the sgRNA leakiness of our CRISPRa system (18) , we speculated that the same strategy might also be applied to the crRNA-tracrRNA mediated CRISPRa. We first verified that dxCas9 could work with our newly designed sgRNA with tail-fused aptamers and functioned better than Cas9 (Supplementary Figure 2) . Then, we proved that the dxCas9 could make the AND gates more symmetrical in response to the two inputs (Supplementary Figure 3) . Notably, previous research has showed that SpCas9 has direct intermolecular interaction with the dual-RNA hybridized fragments (5) . Since many related factors may affect the CRISPR function, we first investigated by varying the length of the RNA hybridization segment. We truncated the WT crRNA and WT tracrRNA simultaneously, making the length of the RNA hybrid segment gradually approach the minimum length for SpCas9 (Figure 2a) . By characterizing them with CRISPRa, we noticed a dramatic decrease of the CRISPRa output when the paired length is shorter than 14 bp (including the two wobble base pairs). Conversely, the 14 bp length could support a similar output level to that from the WT version (Figure 2a, Supplementary Figure 4) . Another question is whether RNase III is necessary for dual-RNA-mediated CRISPRa. For our design, the 3'-end of the crRNA has a terminator sequence, and the 5'-end of the tracrRNA retains a wild-type residual fragment. Both the crRNA and tracrRNA have ends longer than mature dual-RNA processed by RNase III. In previous studies, RNase III was shown to be indispensable for the maturation of crRNA and the immune function of Cas9 (6) . Considering that RNase III has a preference for target double-strand RNA sequence, if RNase III plays a crucial function in our 8 system, then the reprogramming of the crRNA-tracrRNA hybrid sequence may indirectly affect CRISPR function via RNase III. To verify this issue, we introduced our dual-RNA-mediated CRISPRa device into E. coli strains W3110 and its RNase III gene (rnc) knockout strain HT115. Our results show that RNase III is not required for our CRISPRa system. Furthermore, we complemented strain HT115 with a copy of the rnc gene expressed from a plasmid. However, this did not significantly improve the function of CRISPRa (Figure 2b) . This result implies that RNase III will not be a limiting factor for crRNA-tracrRNA reprogramming. Next, we tested the importance of the three bases in the wobble base pairs (G -U) and the bulge in crRNA via two additional experiments. By introducing saturation mutation at the GA site of the WT hybridization sequence, and at the first wobble base pair (G -U) of a functional hybridization sequence U5, we confirmed that changing these compositions will not completely disrupt CRISPRa function, although a few mutations can lead to changes in CRISPR activity (Figure c, d) . To explore potential influencing factors that may affect the activity of these reprogrammed tracrRNA-crRNA pairs, we first manually extracted the following features for each tracrRNA-crRNA sequence: the minimum free energy (MFE) of the crRNA optimal secondary structure (variable 1), change in Gibbs free energy (ΔG) of crRNA-tracrRNA heterodimer binding (variable 2), homology of the crRNA-tracrRNA matching region (variable 3), sequence similarity between the crRNA and the target DNA including the sequence downstream of the PAM site (variable 4) and GC content of the matching region (variable 5). Then we calculated the Pearson correlation coefficients between these features and output fluorescence. shows that only variable 2 and variable 1 have a significant correlation with output fluorescence (p < 0.05). We next built a linear regression model using these two variables to reveal their relationship. Although the R 2 value of the linear model is only 19%, the available data still suggests a trend, which is in line with our understanding of the CRISPR/Cas9 complex formation. The function of CRISPRa tends to be promoted by high affinity between crRNA and tracrRNA (low ΔG for heterodimer binding between crRNA and tracrRNA) and weak secondary structure of crRNA itself (high free energy of the thermodynamic ensemble for crRNA folding) (Figure 2g ). Although pre-crRNA processing is necessary for CRISPR function (6, 23) , the engineering of sgRNA proves that the 3'-end of crRNA near Cas9 is unnecessary for Cas9 function (4). In addition, the extended 5'-end of sgRNA and extended 3'-end of crRNA do not disrupt the activity of CRISPR/Cas9 (22, 24, 25) . Combining the facts that the spacer sequence itself can be reprogrammed to target different DNA sequences, and that the downstream repeat region is also programmable, then the entire crRNA sequence should be programmable. Conversely, any RNA sequence may become crRNA through dual recognition by programmed tracrRNA and target DNA. To verify the above hypothesis, we randomly selected three GA sites on the mRNA encoding red fluorescent protein (RFP), since GA provided the highest function in our saturation mutation test (Figure 2d) . The NGG adjacent to the assumed spacers was deliberately avoided. According to the context sequence adjacent to the GA sites, we designed three corresponding tracrRNAs and three cognate σ 54 -dependent promoters. Each of them has the UAS matching the assumed spacer sequence, which is the upstream mRNA region of the predicted mRNA-tracrRNA hybridization position. If the mRNA of RFP can be hijacked as crRNA, the mRNA should activate the CRISPRa device with dCas9 and corresponding tracrRNAs (Figure 3a) . For each tracrRNA target site on the mRNA, we used the WT tracrRNA, which should not match the RFP mRNA, as a control. Our result shows that, when employing mRNA-paired tracrRNAs, there was a positive linear correlation between the expression level of RFP and GFP for the R2 and R3 sites. This correlation disappeared when using WT tracrRNA (Figure 3b, c) . The three randomly selected tracrRNA target sites gave different CRISPRa efficiencies. The output from site 11 R1 is weak, and those from sites R2 and R3 are relatively stronger (Figure 3b) . For all the sites, when using the corresponding sequences in isolation as a short crRNA, the CRISPRa output intensities obtained were greater than using the whole mRNA (Figure 3d, Supplementary Figure 6 ). We also found that the relationship of CRISPRa strength between different target sites is dependent on the inherent characteristics of mRNA. When the target sequences were fragmented as crRNAs, this strength relationship changes dramatically. It is tempting to speculate that the translation process might cause the above output discrepancy between mRNA and isolated fragments. However, surprisingly, the binding of the CRISPR complex to mRNA did not seem to interfere with RFP translation (Figure 3b) . Conversely, to assess whether the translation process could interfere with the CRISPRa system, a version of mRNA lacking the ribosome binding site (RBS) was used to test the impact of translation on CRISPRa. The experimental results show that the level of RFP was indeed reduced due to the lack of RBS, but the output of mRNA-mediated CRISPRa did not change obviously (Supplementary Figure 7) . In addition, we also found that the length of the mRNA-tracrRNA pairing region can affect the CRISPRa efficiency, and for different targets on mRNA, the optimal length is also different (Supplementary Figure 8) . Although the above results show that, for hijacking mRNA, the CRISPR function of a particular target site is unpredictable, we nevertheless explored some design principles based on two exciting facts we discovered. First, a simplified tracrRNA design can improve the performance of this mRNA monitor on all the sites. Our result shows that when we deleted the redundant 5'-end of tracrRNA, the strength and dynamic range of CRISPRa are greatly improved (Figure 3d, Supplementary Figure 6) . Second, the optimization of tracrRNA did not change the strength pattern of a set of sites on the mRNA. We compared the patterns when CRISPRa is combined with mRNA or fragmented mRNA, with or without tracrRNA optimization. We found that the patterns based on fragmented mRNA and complete mRNA are not similar, however, the patterns based on presence or absence of optimization maintain the similarity when from the mRNA or fragmented mRNA, respectively. This suggests that the inherent characteristics of mRNA such as secondary structure may robustly affect the function of mRNA-mediated CRISPRa (Figure 3e) , and also suggests that site selection and tracrRNA optimization may be independent optimization methods. Finally, we combined the design strategies developed to optimize the mRNA sensor in vivo. To amplify the output signal, we designed a positive feedback loop (PFB) (Figure 3f) . One more σ 54dependent promoter activated by the mRNA was utilized to express additional engineered activator protein for enhancing CRISPRa. The result shows that the PFB circuits could amplify the CRISPRa output and dynamic range based on different mRNA-tracrRNA matching sites (Figure 3g) . The dynamic range increases from 9.6-fold of the original device to 52.2-fold with PFB design through the above engineering methods. Successful hijacking of the RFP mRNA as crRNA raises an interesting question: is the same strategy available for endogenous RNAs transcribed from the genome? To explore this possibility, we chose the ars operon of E. coli as a target, since the promoter P arsR can be induced by sodium arsenite (NaAsO 2 ), a toxic environmental pollutant (26, 27) . Having previously shown that different sites on the mRNA may have different availabilities, we tested some candidate sites on the ars transcripts on artificial circuits first. The arsenic responsive gene cluster (arsRBC) was isolated and cloned into a vector under inducible promoter P lux2 . Four candidate sites were chosen and made into short fragments of mRNA (Figure 4a) . We tested the availability of these sites by using programmed corresponding tracrRNAs and promoters. The results showed that only Ar2 is an available target on the entire ars mRNA. Yet, for the fragmented mRNA, different availabilities of the mRNA sites were revealed, similar to what we have observed in hijacking RFP mRNA. Again, this implies that the characteristics of mRNA itself can strongly affect the function of mRNA-mediated CRISPR (Figure 4b, c) . Next, we used the selected Ar2 site to monitor the activity of the ars operon on the E. coli genome. The reporter circuit and a tracrRNA-Ar2 generator were transformed into E. coli (Figure 4d) . We constitutively induced the expression of dCas9, tracrRNA, and activator in E. coli. When we set a gradient of NaAsO 2 for a group of E. coli cultures, as expected, for tracrRNA-Ar2, an output increase was detected with increasing NaAsO 2 concentration (Figure 4e) . In this case, the output signal is much weaker than that from the artificially expressed mRNA, which may be due to the relatively low concentration of endogenous RNA and the different spatial locations of the transcription in the cell. Finally, we tried to optimize this sensor by simplifying the 5'-end of tracrRNA, and proved that it can improve the performance of the mRNA monitor (Figure 4e, f) . Overall, we confirm that endogenous mRNA can be hijacked by Cas9, and in this way, it allows us to monitor genomic transcriptional activity and connect the cellular gene regulatory network to an artificial actuating or reporting gene circuit. An engineering method that can convert non-crRNA into crRNA undoubtedly has great application potential for nucleic acid sensing. Here, we developed a novel programmable RNA sensor with unique dual recognition characteristics, which theoretically has a higher specificity than the DNA recognition of the CRISPR/Cas9 complex. Compared with other well-known CRISPR RNA sensors such as SHERLOCK, HOLMES and DETECTR that can continuously cleave non-specific singlestranded nucleic acid (28) (29) (30) , Cas9 cleavage lacks such signal amplification effect. In contrast to Cas12a and Cas13, Cas9 can only cleave specific target DNA and then remains bound to it, resulting in insufficient repetitive cleavage of the same target DNA (31) . In order to overcome the above shortcomings of Cas9-directed RNA sensing, we designed an in vitro transcription-based reporting system with a signal amplification effect. We named it the CRISPR-operated Nucleic acid Amputation Notification (CONAN) system. It consists of a reporter DNA named CONAN DNA, fluorogen DFHBI, purified Cas9, and reprogrammed tracrRNA for sensing target RNA in an in vitro T7 expression system (32) (Figure 5a) . The CONAN DNA can express an inactive Broccoli RNA aptamer which is blocked by a 3' end secondary structure. When it senses the target RNA, the Cas9 will be activated and destroy the 3' end secondary structure on the CONAN DNA, which will then continuously transcribe a functional aptamer that can bind to DFHBI, leading to an amplified fluorescent output signal (33) . Due to the strong transcriptional activity of the T7 promoter, once a cleavage event occurs, Broccoli RNA can continue to accumulate in the cell-free system until it reaches a balance with the RNA degradation rate (Figure 5a ). We tested the CONAN sensor using synthetic coronavirus SARS-CoV-2 RNA fragments with corresponding reprogrammed tracrRNAs and Cas9. In this case, we designed a Cas9 cleavage site A between the Broccoli RNA coding sequence and the anti-Broccoli tail in the CONAN DNA. Only when the SARS-CoV-2 RNA fragment A matched the CONAN DNA and corresponding tracrRNA A, did CONAN give a significant report signal (Figure 5b) . Otherwise, even if the CONAN DNA matched the target RNA, there was no false-positive signal. This dual recognition mechanism ensures the high specificity of CONAN detection. Two different protocols, one-step and three-step protocol were designed to further test and optimize the CONAN system (Figure 5c) . Upon sensing the SARS-CoV-2 RNA, we observed fluorescence reported by the CONAN sensor within 10 min (Figure 5d) . The results also indicate that pre-incubation of CONAN DNA with CRISPR/Cas9 could improve the CONAN sensor's output dynamic range. After incubation in the plate reader, we directly photographed the reaction vessel with a cell phone in the fluorescence imaging box. The broccoli RNA concentration in the microwells had accumulated enough to be distinguished by the naked eye and the cell phone camera (Figure 5d) , which indicates that CONAN has the potential to be developed into a portable rapid diagnostic tool. In this work, we investigated the programmability of crRNA-tracrRNA pairing in the CRISPR/Cas9 system based on our previously engineered eukaryote-like CRISPRa device. We systematically revealed the tolerance of SpCas9 protein for reprogrammed sequences of the crRNA-tracrRNA hybridizing region and found that all the nine base pairs necessary for the CRISPR function can be replaced without affecting the CRISPR function significantly. We showed that mismatches can confer orthogonality to crRNA-tracrRNA mediated CRISPRa and design of orthogonal AND logic devices. Interestingly, the mismatch tolerance of different segments of the crRNA-tracrRNA hybridizing region is uneven. The segment close to the spacer is more sensitive to mismatches than the part far away from it. For the RNA duplex between the spacer and bulge, two continuous mismatches or more are sufficient to disrupt the function of CRISPRa. We used a paired crRNA-tracrRNA hybridization library to study the sequence preference of the SpCas9 crRNA-tracrRNA hybridization region. Two key factors, the affinity between crRNA and tracrRNA and the crRNA secondary structure, were identified. It is worth noting that different Cas9 versions may have different tolerance and preference to the reprogrammed crRNA-tracrRNA hybridizing sequence and mismatches. During our previous research, compared with dCas9, dxCas9 shows a lower tolerance for continuous mismatches at specific sites in the crRNA-tracrRNA pairing region (18) . The sequence preference of different Cas9 proteins for the crRNA-tracrRNA hybridizing region may need to be assessed by combining machine learning and highthroughput methods case by case. However, according to various practical scenarios we have covered here (i.e., endogenous RNA hijacking or for the AND gate devices), there are other factors than the aforementioned sequence preferences which may affect the CRISPRa function. Hence it is not easy to predictably select the best RNA target sequence. This reinforces the significance of further research on sequence preference by combining other high-throughput methods and machine learning with only a standardized library. Further studies are still required for further understanding of this issue. Since CRISPR/Cas9 can tolerate extensions of the 5'-end and 3'-end, with enough length, any RNA may become a crRNA by binding to its complementary tracrRNA. Our experiments confirmed this possibility by hijacking of plasmid-transcribed mRNA and endogenous mRNA molecules as crRNAs to trigger CRISPRa function. This is consistent with another recent study (17) . Using the CRISPR/Cas9 system to recognize RNA molecules is an engineering field worth further exploring. However, there are some questions remained to be answered. Different targets on the same mRNA and different tracrRNA lengths on the same target have led to different CRISPRa efficiencies in our experiments. It is unclear whether this was due to RNA secondary structure, translation interference, RNase III, or other unknown factors. In our research, the RBS deletion of the plasmid-transcribed RFP mRNA did not significantly affect CRISPRa output, reflecting the counter-intuitive weak effect of translation on mRNA-mediated CRISPR function. Additionally, whether RNase III can process the mRNA-tracrRNA remains unknown, affecting our assessment about whether the secondary structure of the downstream sequence of the target RNA will affect CRISPR function. We realized that the same system might work in eukaryotes. However, in eukaryotic cells, the localization of endogenous RNA in different regions is an unfavorable factor, affecting the function of endogenous RNA-based CRISPR tools. The tolerance of dCas9 for gRNA sequence diversity also raises an interesting question. Considering that Cas9 binding to non-canonical RNA has recently been reported in bacteria (17), would it be possible for endogenous RNA in eukaryotic cells to coincidentally form a functional CRISPR complex with Cas9, and then cause an off-target effect? In short, exploring the interaction between endogenous RNA and Cas9 in eukaryotes will lead to some new topics that are worth exploring. Finally, we have developed a new cell-free reporter system CONAN with a signal amplification effect on our RNA sensor based on the programmability of crRNA-tracrRNA hybridization. Compared to the recently developed CRISPRi-based in vitro sensing system, CONAN is transcription only and therefore the response time is much shorter upon sensing (17) . Moreover, it only amplifies the output signal when it senses target RNA and has much lower background signal. Thus it will be preferable for broad biosensing applications. This system can be used for CRISPR/Cas9-mediated RNA sensors and for enhancing or amplifying the signal of other types of CRISPR-based RNA sensors, as well as for many other research and application scenarios related to DNA fragmentation. In summary, the programmability of crRNA-tracrRNA hybridization and the RNA recognition capability of the CRISPR/Cas9 system opens up a significantly broader picture of CRISPR/Cas9 engineering. For example, this programmable system may be utilized to sense RNA in vivo or in vitro, to build RNA editors, to reconstruct the topology of the genetic regulatory network, to mark transcripts in situ, or to build complex cellular computing circuits. Our findings have elevated our knowledge of CRISPR systems and will expand current CRISPR technologies in the future. Unless otherwise specified, all the assays were performed in E. coli strain MC1061ΔpspF. The E. coli MC1061ΔpspF was generated through P1 phage transduction, using E. coli strain BW25113ΔpspF739::kan from the Keio collection as the donor strain (35) . For circuit construction, E. coli TOP10 and E. coli DH5α were both employed in this study. We used standard molecular biology protocols in this study for circuit construction. All plasmids and their structures and sequences are listed in Supplementary Table 3 All the genetic parts of the dCas9 generator, dxCas9 generator, activator generator, inducible promoters, and sgRNA with aptamers at the tetraloop come from our previous published study (18) . The new versions of sgRNA were synthesized by annealing of oligonucleotides (j5 protocol (37) ). PCR was used to split sgRNA to crRNA and tracrRNA, introduce mutations into gRNAs, and change the spacer sequence of crRNAs. All the plasmids have been sequenced to confirm the sequence (Source Bioscience), except for the CONAN DNA, which cannot be sequenced because it contains a long reverse palindrome sequence, and it has been checked by HpyAV restriction mapping. The in vitro CONAN assays were performed in a 10 L reaction volume using a previously To prepare for cell phone imaging at the end of CONAN assays, the microplate was placed onto the surface of a Safe Imager (S37102, Invitrogen) blue-light trans-illuminator and was covered with an amber filter in a dark environment. A cell phone (iPhone 11) was used to acquire the images with the build-in night mode and 3 s auto exposure. No additional adjustments were made to the images. For the characterization of orthogonality and AND logic gates (Figure 1f, 1g) Random sequences were generated by online tool (38) , then the sense strand and antisense strand sequences were used for the paired segments of crRNA and tracrRNA, respectively. Among them, tracrRNA retains the bulge structure and the G for the wobble base pairs. We use Excel 2016 to check whether there is an NGG in crRNA next to the spacer and exclude these sequences. Sticky ends compatible with the expression vector were designed to ensure that the DNA annealing products can be directly used for ligation. The randomized sequences for every reprogramed crRNA and tracrRNA were ordered as complementary DNA oligonucleotides (Merck), annealed and cloned into the respective backbone plasmids. The annealing was carried out in 96-well PCR plates by mixing 2 µL of each oligonucleotide at 100 µM with 2 µL of 10x T4 DNA Ligase Reaction Buffer (NEB) and 14 µL of water. The plate was then incubated at 95 ºC for 5 min in a thermocycler and allowed to cool down within the machine until reaching room temperature (c.a. 20 min). The annealed oligonucleotides produce overhangs compatible with the digested destination plasmids. The crRNA duplexes were cloned into the plasmid pLY257, previously digested with BbsI-HF (NEB), and the tracrRNA duplexes were cloned into the plasmid pLY258, previously digested with BsaI-HFv2 (NEB). Backbone plasmids were purified from gel after digestion using the Monarch DNA Gel Extraction Kit (NEB). Ligations were carried out in 96-well PCR plates by mixing 0.2 µL of digested plasmid The Pearson correlation analysis was conducted with the R package "ggcorrplot", and the regression model was built with the MATLAB 2015b function "regress". Data availability: All data in the main text and the supplementary materials are available upon reasonable request. Structures of Cas9 endonucleases reveal RNA-mediated conformational activation Structural basis of PAM-dependent target DNA recognition by the Cas9 endonuclease Classification and evolution of type II CRISPR-Cas systems CRISPR-Cas9 Structures and Mechanisms Crystal Structure of Cas9 in Complex with Guide RNA and Target DNA CRISPR RNA maturation by trans-encoded small RNA and host factor RNase III Biogenesis pathways of RNA guides in archaeal and bacterial CRISPR-Cas adaptive immunity The tracrRNA and Cas9 families of type II CRISPR-Cas immunity systems CRISPR-Cas immunity in prokaryotes CRISPR provides acquired resistance against viruses in prokaryotes The new frontier of genome engineering with CRISPR-Cas9 Structural roles of guide RNAs in the nuclease activity of Cas9 endonuclease A Programmable Dual-RNA-Guided DNA Endonuclease in Adaptive Bacterial Immunity Guide RNA Functional Modules Direct Cas9 Activity and Orthogonality Simultaneous repression of multiple bacterial genes using nonrepetitive extra-long sgRNA arrays A scoutRNA is required for some type V CRISPR-Cas Systems Noncanonical crRNAs derived from host transcripts enable multiplexable RNA detection by Cas9 Engineered CRISPRa enables programmable eukaryote-like gene activation in bacteria A novel eukaryote-like CRISPR activation tool in bacteria: features and capabilities Recent advances and opportunities in synthetic logic gates engineering in living cells Orthogonality and burdens of heterologous AND gate gene circuits in E. coli Riboregulated toehold-gated gRNA for programmable CRISPR-Cas9 function crRNA and tracrRNA guide Cas9-mediated DNA interference in Streptococcus thermophilus Aptazyme-embedded guide RNAs enable ligand-responsive genome editing and transcriptional activation Directing cellular information flow via CRISPR signal conductors Cascaded amplifying circuits enable ultrasensitive cellular sensors for toxic metals The chromosomal arsR gene of Escherichia coli encodes a trans-acting metalloregulatory protein Nucleic acid detection with CRISPR-Cas13a/C2c2 CRISPR-Cas12a-assisted nucleic acid detection CRISPR-Cas12a target binding unleashes indiscriminate single-stranded DNase activity DNA interrogation by the CRISPR RNA-guided endonuclease Cas9 TXO: Transcription-Only genetic circuits as a novel cell-free approach for Synthetic Biology Broccoli: rapid selection of an RNA mimic of green fluorescent protein by fluorescence-based selection and directed evolution Mfold web server for nucleic acid folding and hybridization prediction Construction of Escherichia coli K-12 in-frame, single-gene knockout mutants: the Keio collection Engineering BioBrick vectors from BioBrick parts Keasling, j5 DNA assembly design automation software The sequence manipulation suite: JavaScript programs for analyzing and formatting protein and DNA sequences We thank Dr Jennifer Tullet (University of Kent) for supplying the W3110 and HT115 E. coli strains used in this study. This work was supported by the UK Research and Innovation Future Leaders The authors declare no competing interests.