Uncertainty-aware and interpretable evaluation of Cas9-gRNA and Cas12a-gRNA specificity for fully matched and partially mismatched targets with Deep Kernel Learning.

The choice of guide RNA (gRNA) for CRISPR-based gene targeting is an essential step in gene editing applications, but the prediction of gRNA specificity remains challenging. Lack of transparency and focus on point estimates of efficiency disregarding the information on possible error sources in the model limit the power of existing Deep Learning-based methods. To overcome these problems, we present a new approach, a hybrid of Capsule Networks and Gaussian Processes. Our method predicts the cleavage efficiency of a gRNA with a corresponding confidence interval, which allows the user to incorporate information regarding possible model errors into the experimental design. We provide the first utilization of uncertainty estimation in computational gRNA design, which is a critical step toward accurate decision-making for future CRISPR applications. The proposed solution demonstrates acceptable confidence intervals for most test sets and shows regression quality similar to existing models. We introduce a set of criteria for gRNA selection based on off-target cleavage efficiency and its variance and present a collection of pre-computed gRNAs for human chromosome 22. Using Neural Network Interpretation methods, we show that our model rediscovers an established biological factor underlying cleavage efficiency, the importance of the seed region in gRNA.

Prespacers formed during primed adaptation associate with the Cas1-Cas2 adaptation complex and the Cas3 interference nuclease-helicase.

For Type I CRISPR-Cas systems, a mode of CRISPR adaptation named priming has been described. Priming allows specific and highly efficient acquisition of new spacers from DNA recognized (primed) by the Cascade-crRNA (CRISPR RNA) effector complex. Recognition of the priming protospacer by Cascade-crRNA serves as a signal for engaging the Cas3 nuclease-helicase required for both interference and primed adaptation, suggesting the existence of a primed adaptation complex (PAC) containing the Cas1-Cas2 adaptation integrase and Cas3. To detect this complex in vivo, we here performed chromatin immunoprecipitation with Cas3-specific and Cas1-specific antibodies using cells undergoing primed adaptation. We found that prespacers are bound by both Cas1 (presumably, as part of the Cas1-Cas2 integrase) and Cas3, implying direct physical association of the interference and adaptation machineries as part of PAC.

Systematic analysis of Type I-E Escherichia coli CRISPR-Cas PAM sequences ability to promote interference and primed adaptation.

CRISPR interference occurs when a protospacer recognized by the CRISPR RNA is destroyed by Cas effectors. In Type I CRISPR-Cas systems, protospacer recognition can lead to «primed adaptation¼ - acquisition of new spacers from in cis located sequences. Type I CRISPR-Cas systems require the presence of a trinucleotide protospacer adjacent motif (PAM) for efficient interference. Here, we investigated the ability of each of 64 possible trinucleotides located at the PAM position to induce CRISPR interference and primed adaptation by the Escherichia coli Type I-E CRISPR-Cas system. We observed clear separation of PAM variants into three groups: those unable to cause interference, those that support rapid interference and those that lead to reduced interference that occurs over extended periods of time. PAM variants unable to support interference also did not support primed adaptation; those that supported rapid interference led to no or low levels of adaptation, while those that caused attenuated levels of interference consistently led to highest levels of adaptation. The results suggest that primed adaptation is fueled by the products of CRISPR interference. Extended over time interference with targets containing «attenuated¼ PAM variants provides a continuous source of new spacers leading to high overall level of spacer acquisition.

CRISPR-Cas adaptation in Escherichia coli requires RecBCD helicase but not nuclease activity, is independent of homologous recombination, and is antagonized by 5' ssDNA exonucleases.

Prokaryotic adaptive immunity is established against mobile genetic elements (MGEs) by 'naïve adaptation' when DNA fragments from a newly encountered MGE are integrated into CRISPR-Cas systems. In Escherichia coli, DNA integration catalyzed by Cas1-Cas2 integrase is well understood in mechanistic and structural detail but much less is known about events prior to integration that generate DNA for capture by Cas1-Cas2. Naïve adaptation in E. coli is thought to depend on the DNA helicase-nuclease RecBCD for generating DNA fragments for capture by Cas1-Cas2. The genetics presented here show that naïve adaptation does not require RecBCD nuclease activity but that helicase activity may be important. RecA loading by RecBCD inhibits adaptation explaining previously observed adaptation phenotypes that implicated RecBCD nuclease activity. Genetic analysis of other E. coli nucleases and naïve adaptation revealed that 5' ssDNA tailed DNA molecules promote new spacer acquisition. We show that purified E. coli Cas1-Cas2 complex binds to and nicks 5' ssDNA tailed duplexes and propose that E. coli Cas1-Cas2 nuclease activity on such DNA structures supports naïve adaptation.

The action of Escherichia coli CRISPR-Cas system on lytic bacteriophages with different lifestyles and development strategies.

CRISPR-Cas systems provide prokaryotes with adaptive defense against bacteriophage infections. Given an enormous variety of strategies used by phages to overcome their hosts, one can expect that the efficiency of protective action of CRISPR-Cas systems against different viruses should vary. Here, we created a collection of Escherichia coli strains with type I-E CRISPR-Cas system targeting various positions in the genomes of bacteriophages λ, T5, T7, T4 and R1-37 and investigated the ability of these strains to resist the infection and acquire additional CRISPR spacers from the infecting phage. We find that the efficiency of CRISPR-Cas targeting by the host is determined by phage life style, the positions of the targeted protospacer within the genome, and the state of phage DNA. The results also suggest that during infection by lytic phages that are susceptible to CRISPR interference, CRISPR-Cas does not act as a true immunity system that saves the infected cell but rather enforces an abortive infection pathway leading to infected cell death with no phage progeny release.

Spacer-length DNA intermediates are associated with Cas1 in cells undergoing primed CRISPR adaptation.

During primed CRISPR adaptation spacers are preferentially selected from DNA recognized by CRISPR interference machinery, which in the case of Type I CRISPR-Cas systems consists of CRISPR RNA (crRNA) bound effector Cascade complex that locates complementary targets, and Cas3 executor nuclease/helicase. A complex of Cas1 and Cas2 proteins is capable of inserting new spacers in the CRISPR array. Here, we show that in Escherichia coli cells undergoing primed adaptation, spacer-sized fragments of foreign DNA are associated with Cas1. Based on sensitivity to digestion with nucleases, the associated DNA is not in a standard double-stranded state. Spacer-sized fragments are cut from one strand of foreign DNA in Cas1- and Cas3-dependent manner. These fragments are generated from much longer S1-nuclease sensitive fragments of foreign DNA that require Cas3 for their production. We propose that in the course of CRISPR interference Cas3 generates fragments of foreign DNA that are recognized by the Cas1-Cas2 adaptation complex, which excises spacer-sized fragments and channels them for insertion into CRISPR array.

Highly efficient primed spacer acquisition from targets destroyed by the Escherichia coli type I-E CRISPR-Cas interfering complex.

Prokaryotic clustered regularly interspaced short palindromic repeat (CRISPR)-CRISPR associated (Cas) immunity relies on adaptive acquisition of spacers-short fragments of foreign DNA. For the type I-E CRISPR-Cas system from Escherichia coli, efficient "primed" adaptation requires Cas effector proteins and a CRISPR RNA (crRNA) whose spacer partially matches a segment (protospacer) in target DNA. Primed adaptation leads to selective acquisition of additional spacers from DNA molecules recognized by the effector-crRNA complex. When the crRNA spacer fully matches a protospacer, CRISPR interference-that is, target destruction without acquisition of additional spacers-is observed. We show here that when the rate of degradation of DNA with fully and partially matching crRNA targets is made equal, fully matching protospacers stimulate primed adaptation much more efficiently than partially matching ones. The result indicates that different functional outcomes of CRISPR-Cas response to two kinds of protospacers are not caused by different structures formed by the effector-crRNA complex but are due to the more rapid destruction of targets with fully matching protospacers.

Altered stoichiometry Escherichia coli Cascade complexes with shortened CRISPR RNA spacers are capable of interference and primed adaptation.

The Escherichia coli type I-E CRISPR-Cas system Cascade effector is a multisubunit complex that binds CRISPR RNA (crRNA). Through its 32-nucleotide spacer sequence, Cascade-bound crRNA recognizes protospacers in foreign DNA, causing its destruction during CRISPR interference or acquisition of additional spacers in CRISPR array during primed CRISPR adaptation. Within Cascade, the crRNA spacer interacts with a hexamer of Cas7 subunits. We show that crRNAs with a spacer length reduced to 14 nucleotides cause primed adaptation, while crRNAs with spacer lengths of more than 20 nucleotides cause both primed adaptation and target interference in vivo Shortened crRNAs assemble into altered-stoichiometry Cascade effector complexes containing less than the normal amount of Cas7 subunits. The results show that Cascade assembly is driven by crRNA and suggest that multisubunit type I CRISPR effectors may have evolved from much simpler ancestral complexes.

Dynamics of Escherichia coli type I-E CRISPR spacers over 42 000 years.

CRISPR-Cas are nucleic acid-based prokaryotic immune systems. CRISPR arrays accumulate spacers from foreign DNA and provide resistance to mobile genetic elements containing identical or similar sequences. Thus, the set of spacers present in a given bacterium can be regarded as a record of encounters of its ancestors with genetic invaders. Such records should be specific for different lineages and change with time, as earlier acquired spacers get obsolete and are lost. Here, we studied type I-E CRISPR spacers of Escherichia coli from extinct pachyderm. We find that many spacers recovered from intestines of a 42 000-year-old mammoth match spacers of present-day E. coli. Present-day CRISPR arrays can be reconstructed from palaeo sequences, indicating that the order of spacers has also been preserved. The results suggest that E. coli CRISPR arrays were not subject to intensive change through adaptive acquisition during this time.

The Digital Ageing Atlas: integrating the diversity of age-related changes into a unified resource.

Multiple studies characterizing the human ageing phenotype have been conducted for decades. However, there is no centralized resource in which data on multiple age-related changes are collated. Currently, researchers must consult several sources, including primary publications, in order to obtain age-related data at various levels. To address this and facilitate integrative, system-level studies of ageing we developed the Digital Ageing Atlas (DAA). The DAA is a one-stop collection of human age-related data covering different biological levels (molecular, cellular, physiological, psychological and pathological) that is freely available online (http://ageing-map.org/). Each of the >3000 age-related changes is associated with a specific tissue and has its own page displaying a variety of information, including at least one reference. Age-related changes can also be linked to each other in hierarchical trees to represent different types of relationships. In addition, we developed an intuitive and user-friendly interface that allows searching, browsing and retrieving information in an integrated and interactive fashion. Overall, the DAA offers a new approach to systemizing ageing resources, providing a manually-curated and readily accessible source of age-related changes.

Foreign DNA acquisition by the I-F CRISPR-Cas system requires all components of the interference machinery.

CRISPR immunity depends on acquisition of fragments of foreign DNA into CRISPR arrays. For type I-E CRISPR-Cas systems two modes of spacer acquisition, naïve and primed adaptation, were described. Naïve adaptation requires just two most conserved Cas1 and Cas2 proteins; it leads to spacer acquisition from both foreign and bacterial DNA and results in multiple spacers incapable of immune response. Primed adaptation requires all Cas proteins and a CRISPR RNA recognizing a partially matching target. It leads to selective acquisition of spacers from DNA molecules recognized by priming CRISPR RNA, with most spacers capable of protecting the host. Here, we studied spacer acquisition by a type I-F CRISPR-Cas system. We observe both naïve and primed adaptation. Both processes require not just Cas1 and Cas2, but also intact Csy complex and CRISPR RNA. Primed adaptation shows a gradient of acquisition efficiency as a function of distance from the priming site and a strand bias that is consistent with existence of single-stranded adaption intermediates. The results provide new insights into the mechanism of spacer acquisition and illustrate surprising mechanistic diversity of related CRISPR-Cas systems.

The Cas6e ribonuclease is not required for interference and adaptation by the E. coli type I-E CRISPR-Cas system.

CRISPR-Cas are small RNA-based adaptive prokaryotic immunity systems protecting cells from foreign DNA or RNA. Type I CRISPR-Cas systems are composed of a multiprotein complex (Cascade) that, when bound to CRISPR RNA (crRNA), can recognize double-stranded DNA targets and recruit the Cas3 nuclease to destroy target-containing DNA. In the Escherichia coli type I-E CRISPR-Cas system, crRNAs are generated upon transcription of CRISPR arrays consisting of multiple palindromic repeats and intervening spacers through the function of Cas6e endoribonuclease, which cleaves at specific positions of repeat sequences of the CRISPR array transcript. Cas6e is also a component of Cascade. Here, we show that when mature unit-sized crRNAs are provided in a Cas6e-independent manner by transcription termination, the CRISPR-Cas system can function without Cas6e. The results should allow facile interrogation of various targets by type I-E CRISPR-Cas system in E. coli using unit-sized crRNAs generated by transcription.

Pervasive generation of oppositely oriented spacers during CRISPR adaptation.

During the process of prokaryotic CRISPR adaptation, a copy of a segment of foreign deoxyribonucleic acid referred to as protospacer is added to the CRISPR cassette and becomes a spacer. When a protospacer contains a neighboring target interference motif, the specific small CRISPR ribonucleic acid (crRNA) transcribed from expanded CRISPR cassette can protect a prokaryotic cell from virus infection or plasmid transformation and conjugation. We show that in Escherichia coli, a vast majority of plasmid protospacers generate spacers integrated in CRISPR cassette in two opposing orientations, leading to frequent appearance of complementary spacer pairs in a population of cells that underwent CRISPR adaptation. When a protospacer contains a spacer acquisition motif AAG, spacer orientation that generates functional protective crRNA is strongly preferred. All other protospacers give rise to spacers oriented in both ways at comparable frequencies. This phenomenon increases the repertoire of available spacers and should make it more likely that a protective crRNA is formed as a result of CRISPR adaptation.

High-throughput analysis of type I-E CRISPR/Cas spacer acquisition in E. coli.

In Escherichia coli, the acquisition of new CRISPR spacers is strongly stimulated by a priming interaction between a spacer in CRISPR RNA and a protospacer in foreign DNA. Priming also leads to a pronounced bias in DNA strand from which new spacers are selected. Here, ca. 200,000 spacers acquired during E. coli type I-E CRISPR/Cas-driven plasmid elimination were analyzed. Analysis of positions of plasmid protospacers from which newly acquired spacers have been derived is inconsistent with spacer acquisition machinery sliding along the target DNA as the primary mechanism responsible for strand bias during primed spacer acquisition. Most protospacers that served as donors of newly acquired spacers during primed spacer acquisition had an AAG protospacer adjacent motif, PAM. Yet, the introduction of multiple AAG sequences in the target DNA had no effect on the choice of protospacers used for adaptation, which again is inconsistent with the sliding mechanism. Despite a strong preference for an AAG PAM during CRISPR adaptation, the AAG (and CTT) triplets do not appear to be avoided in known E. coli phages. Likewise, PAM sequences are not avoided in Streptococcus thermophilus phages, indicating that CRISPR/Cas systems may not have been a strong factor in shaping host-virus interactions.

PRE-mediated bypass of two Su(Hw) insulators targets PcG proteins to a downstream promoter.

Drosophila Polycomb group response elements (PRE) silence neighboring genes, but silencing can be blocked by one copy of the Su(Hw) insulator element. We show here that Polycomb group (PcG) proteins can spread from a PRE in the flanking chromatin region and that PRE blocking depends on a physical barrier established by the insulator to PcG protein spreading. On the other hand, PRE-mediated silencing can bypass two Su(Hw) insulators to repress a downstream reporter gene. Strikingly, insulator bypass involves targeting of PcG proteins to the downstream promoter, while they are completely excluded from the intervening insulated domain. This shows that PRE-dependent silencing is compatible with looping of the PRE in order to bring PcG proteins in contact with the promoter and does not require the coating of the whole chromatin domain between PRE and promoter.

Interaction between a pair of gypsy insulators or between heterologous gypsy and Wari insulators modulates Flp site-specific recombination in Drosophila melanogaster.

Chromatin insulators block the action of transcriptional enhancers when interposed between an enhancer and a promoter. An Flp technology was used to examine interactions between Drosophila gypsy and Wari insulators in somatic and germ cells. The gypsy insulator consists of 12 binding sites for the Su(Hw) protein, while the endogenous Wari insulator, located on the 3' side of the white gene, is independent from the Su(Hw) protein. Insertion of the gypsy but not Wari insulator between FRT sites strongly blocks recombination between Flp dimers bound to FRT sites located on the same chromatid (recombination in cis) or in sister chromatids (unequal recombination in trans). At the same time, the interaction between Wari and gypsy insulators regulates the efficiency of Flp-mediated recombination. Thus, insulators may have a role in controlling interactions between distantly located protein complexes (not only those involved in transcriptional gene regulation) on the same chromosome or on sister chromatids in somatic and germ cells. We have also found that the frequency of Flp-mediated recombination between FRT sites is strongly dependent on the relative orientation of gypsy insulators. Taken together, our results indicate that the interactions between insulators can be visualized by Flp technology and that insulators may be involved in blocking undesirable interactions between proteins at the two-chromatid phase of the cell cycle.

SCRAMBLER: A Tool for De Novo CRISPR Array Reconstruction and Its Application for Analysis of the Structure of Prokaryotic Populations.

CRISPR arrays are prokaryotic genomic loci consisting of repeat sequences alternating with unique spacers acquired from foreign nucleic acids. As one of the fastest-evolving parts of the genome, CRISPR arrays can be used to differentiate closely related prokaryotic lineages and track individual strains in prokaryotic communities. However, the assembly of full-length CRISPR arrays sequences remains a problem. Here, we developed SCRAMBLER, a tool that includes several pipelines for assembling CRISPR arrays from high-throughput short-read sequencing data. We assessed its performance with model data sets (Escherichia coli strains containing different CRISPR arrays and imitating prokaryotic communities of different complexities) and intestinal microbiomes of extant and extinct pachyderms. Evaluation of SCRAMBLER's performance using model data sets demonstrated its ability to assemble CRISPR arrays correctly from reads containing pairs of spacers, yielding a precision rate of >80% and a recall rate of 60-85% when checked against ground-truth data. Likewise, SCRAMBLER successfully assembled CRISPR arrays from the environmental samples, as attested by their matching with database entries. SCRAMBLER, an open-source software (github.com/biolab-tools/SCRAMBLER), can facilitate analysis of the composition and dynamics of CRISPR arrays in complex communities.

Zeste can facilitate long-range enhancer-promoter communication and insulator bypass in Drosophila melanogaster.

The looping model of enhancer-promoter interactions predicts that these specific long-range interactions are supported by a certain class of proteins. In particular, the Drosophila transcription factor Zeste was hypothesized to facilitate long-distance associations between enhancers and promoters. We have re-examined the role of Zeste in supporting long-range interactions between an enhancer and a promoter using the white gene as a model system. The results show that Zeste binds to the upstream white promoter region and the enhancer that is responsible for white activation in the eyes. We have confirmed the previous finding that Zeste is not required for the activity of the eye enhancer and the promoter when they are located in close proximity to each other. However, inactivation of Zeste markedly affects the enhancer-promoter communication in transgenes when the eye enhancer and the white promoter are separated by a 3-kb spacer or the yellow gene. Zeste is also required for insulator bypass by the eye enhancer. Taken together, these results show that Zeste can support specific long-range interactions between enhancers and promoters.

Red flag on the white reporter: a versatile insulator abuts the white gene in Drosophila and is omnipresent in mini-white constructs.

Much of the research on insulators in Drosophila has been done with transgenic constructs using the white gene (mini-white) as reporter. Hereby we report that the sequence between the white and CG32795 genes in Drosophila melanogaster contains an insulator of a novel kind. Its functional core is within a 368 bp segment almost contiguous to the white 3'UTR, hence we name it as Wari (white-abutting resident insulator). Though Wari contains no binding sites for known insulator proteins and does not require Su(Hw) or Mod(mdg4) for its activity, it can equally well interact with another copy of Wari and with unrelated Su(Hw)-dependent insulators, gypsy or 1A2. In its natural downstream position, Wari reinforces enhancer blocking by any of the three insulators placed between the enhancer and the promoter; again, Wari-Wari, Wari-gypsy or 1A2-Wari pairing results in mutual neutralization (insulator bypass) when they precede the promoter. The distressing issue is that this element hides in all mini-white constructs employed worldwide to study various insulators and other regulatory elements as well as long-range genomic interactions, and its versatile effects could have seriously influenced the results and conclusions of many works.

Mapping CRISPR spaceromes reveals vast host-specific viromes of prokaryotes.

CRISPR arrays contain spacers, some of which are homologous to genome segments of viruses and other parasitic genetic elements and are employed as portion of guide RNAs to recognize and specifically inactivate the target genomes. However, the fraction of the spacers in sequenced CRISPR arrays that reliably match protospacer sequences in genomic databases is small, leaving the question of the origin(s) open for the great majority of the spacers. Here, we extend the spacer analysis by examining the distribution of partial matches (matching k-mers) between spacers and genomes of viruses infecting the given host as well as the host genomes themselves. The results indicate that most of the spacers originate from the host-specific viromes, whereas self-targeting is strongly selected against. However, we present evidence that the vast majority of the viruses comprising the viromes currently remain unknown although they are likely to be related to identified viruses.