Bacteria and archaea evolved CRISPR as part of their adaptive immune system to protect themselves from invading viruses and foreign plasmids. The defence system relies on small, non-coding RNA molecules (CRISPR RNAs/guide RNA) that in association with a CRISPR associated (Cas) protein silence foreign sequences by means of cleavage. Twenty nucleotides at the 5’ end of the CRISPR RNA, corresponding to the protospacer region, direct the Cas9 protein to a specific sequence within the DNA using predictable Watson and Crick base pairing between the target DNA and the guide RNA, triggering enzymatic cleavage of the foreign DNA. Essential for cleavage is the presence of a sequence motif immediately downstream of the target region, known as the protospacer-adjacent motif (PAM),the PAM sequence in S. pyogenes (the most widely used system) is NGG. Because a sequence match also exists between the guide RNA and the host’s DNA that encodes it, it is vital that the guide RNA identifies bona fide targets without attacking the host chromosome. Discrimination between the target and host protospacer is achieved by the Cas effector recognition of a PAM sequence that initiates binding and cleavage of the foreign DNA. Changing the DNA target site that is specified by the 20 nucleotides at the 5’ end of the guide RNA redirects the system to target a different DNA sequence, making it a highly versatile and programmable tool. In 2012, the CRISPR system was programmed to cleave specific DNA sites in vitro and since then, the system has been widely used in many gene modification applications.
CRISPR allow so much more than just substituting nucletodies. Like you say, CRISPR is a much more precise system than restriction enzymes, the guide RNA for CRISPR systems is typically around 20 nucleotides long so it can theoretically be used to target a unique site in the genome (4^20 = 10^12, so only the recognition site is expected to feature once 1 in 10^12 bases, N.B. the human genome is only 3x10^9 base pairs in size).
There are several options using the CRISPR with an active Cas9 protein, the first is to create a gene knockout, which removes the function of a gene. Gene knockouts can be performed by targeting the CRISPR-Cas9 system to the start of the gene and cleaving both strands of the DNA to generate a double stranded break which will be repaired most likely by the cellular repair pathway non homologous end joining (NHEJ). NHEJ results in insertions or deletions and so will most likely result in a frameshift mutation which alters a large amount of the amino acid sequence in the gene or introduces an early stop codon, so the protein produced is non functional. Alternatively a conserved residue (e.g. a catalytically important residue) in the protein can be targeted and mutation of this can remove the catalytic function or recognition domain in an enzyme or protein. NHEJ can also be used to mutate DNA that codes for RNA products. Additionally, the CRISPR system can be targeted to regions outside of the coding sequence such as the promoter or enhancers and silencers/operators (depending on whether prokaryotic or eukaryotic), which can also remove the functionality of the gene.
The CRISPR system can also be introduced to the cell together with a single-stranded or double-stranded DNA template to make very small and precise changes to the DNA (e.g. single base mutations), or insert DNA sequences up to thousands of bases in length so that entire genes can be incorporated into the target region. This more precise editing method also involves the double stranded break formation by the cas9 protein, but relies on different cellular repair pathways: homology directed repair, the most common form of which is homologous recombination. Importantly, the template DNA must have homologous sequences to the DNA flanking either side of the double stranded break. In the presence of the correct exogenous DNA template, homology directed repair can occur in which the exogenous template DNA is incorporated into the target DNA in exchange for the damaged DNA.
Other use of CRIPSR include using a catalytically inactive cas9 protein (dead Cas9). This system is unable to cleave DNA, but is still able to target specific sites in the DNA. dead Cas9 can be fused to or recruit effector proteins for regulating gene expression at the transcriptional level, as well as guiding enzymes capable of precise base editing without introducing double stranded breaks. The dead cas9 system can be fused to transcription repressor domains such as the Krüppel-associated box (KRAB) or transcriptional acitvators such as VP64 for effective and specific repression or activation of gene expression at sites that don't contain recongion sequences for these effectors. dead Cas9 can also be used to recruit epigenetic modifitying enzymes such as methylases and acetylases to regulate gene expression this way. Additionally, this system can be fused to DNA base editors, which are enzymes created by directed evolution that are capable of causing efficient and specific point mutations in DNA and RNA (this is much more specific and effective than point mutations using homology directed repair), although RNA editing required using cas13.
Gene editing can be used to repair disease causing mutations, introduce new genes with novel functions into organisms' genomes, transiently alter gene expression to treat diseases; create knockout, knockdown and knock in strains for studying the effects of genetic factors on phenotype and even more.
In contrast, restriction enzymes typically have recognition site of 6 nucleotides and so they will cut at many orders of magnitude more sites than CRISPR will (4^6 = 4096, so 1 in 4096 nucleotides are expected to be a cut site for restriction enzymes) so they are not suitable for specific editing of genomes as they will cause non specific fragmentation of genomic DNA, which will kill the organism, so CRISPR is a much better enzyme for editing genomes. Instead, restriction enzymes are used in molecular cloning to generate sticky overhangs (or blunt ends) for the assembly of DNA molecules to generate recombinant DNA.
Although the PAM appears to limit the number of target sites that the CRISPR system can be targeted to (around 1 in every 16), through directed evolution, a Cas9 variant with increased PAM compatibility has been developed. This new Cas9 is capable of recognizing PAM sequences including NG, GAA, and GAT, this makes it more versatile and greatly expands the number of targets available for site-sensitive genome editing applications.
Whilst CRISPR based gene editing is very exciting, a lot of work is still needed to: increase the specificity of the system as it causes numerous off site mutations, which can be lethal to cells; to increase the effectiveness of delivery of the system in vivo to large organisms such as humans; to reduce the immunogenicity of the system (it can trigger an immune response in humans) so that it is safe to use; as well as increase our understanding of the affect of making specific mutations in humans (e.g. which mutations can treat diseases safely).
References/further reading:
Jinek, M. et al. A programmable dual-RNA-guided DNA endonuclease in adaptive bacterial immunity. Science (80-. ). 337, 816–821 (2012).
Repurposing CRISPR as an RNA-Guided Platform for Sequence-Specific Control of Gene Expression. Cell 152, 1173–1183 (2013).
Liu, X. S. et al. Editing DNA Methylation in the Mammalian Genome. Cell 167, 233–247.e17 (2016).
Zalatan, J. G. et al. Engineering complex synthetic transcriptional programs with CRISPR RNA scaffolds. Cell 160, 339–50 (2015).
Hu, J. H. et al. Evolved Cas9 variants with broad PAM compatibility and high DNA specificity. Nature (2018). doi:10.1038/nature26155
Cox, D. B. T. et al. RNA editing with CRISPR-Cas13. Science (80-. ). 358, 1019–1027 (2017).