Editing the bases

Model genetic diseases in mice and macaques were cured by the method of base editing

Ekaterina Gracheva, "Elements"

Genetic diseases often occur due to the replacement of just one nucleotide in the genome. Methods based on the CRISPR/Cas9 system that can correct either cytosine for thymine or adenine for guanine in the right position – the so–called base editing - appeared just a few years ago, but are already being used to correct mutations in laboratory animals. Recently, a group of American researchers managed to correct a mutation in the β-globin gene responsible for the development of sickle cell anemia, and another international group of scientists managed to create a mutant version of the PCSK9 gene, which reduces the level of low–density lipoproteins. Both works use upgraded versions of the base editing method. It is not yet known whether this approach will be used in clinical practice, but if it happens, it will help many people to return to normal life.

Approximately half of all variants of the human genome that are associated with pathological conditions occur when only one nucleotide in the DNA sequence is changed (this can occur during replacement, insertion or deletion). Often such changes in the genome are already observed in parents, but manifest themselves in the form of pathology in children if gametes carrying changes are encountered during fertilization. Also, single-nucleotide substitutions can occur spontaneously during the formation of germ cells. Such a change can disrupt the work of one protein, which is enough to develop a serious disease (for example, sickle cell anemia), or can contribute to multifactorial diseases (atherosclerosis, Alzheimer's disease, and many others). The creation of gene therapy methods is primarily aimed at correcting such changes.

Since scientists have mastered the use of the bacterial "immune system" – CRISPR/Cas9 – to make changes to the genetic information in cells, they have the opportunity to model diseases by creating mutations in the genome regions corresponding to mutations in patients, and use them, for example, to select the appropriate drug therapy. At the same time, these mutations can be purposefully corrected in the body of model animals, creating a basis for future genetic therapy of these diseases.

Sickle cell anemia is caused by a single nucleotide substitution in the β-globin gene (beta hemoglobin subunit, HBB), which leads to the replacement of glutamate (GAG) with valine (GTG). Molecules of such an abnormal form of hemoglobin (HbS) "stick together" into fibers, changing the shape of red blood cells, reducing their ability to carry oxygen, which leads to damage and blockage of blood vessels. The mechanism of this process is very well studied, so sickle cell anemia is practically a model disease for gene therapy. Now this disease can be cured only by bone marrow transplantation. This procedure requires, firstly, finding a suitable donor, and secondly, a huge burden on the patient's body.

The scientific literature has already described relatively successful attempts to edit mutation in hematopoietic stem cells of patients using nucleases with "zinc fingers": normal forms of hemoglobin were formed in those who underwent therapy (M. D. Hoban et al., 2015. Correction of the sickle cell disease mutation in human hematopoietic stem/progenitor cells). In 2016, researchers from the USA presented a paper where the same editing was carried out using the CRISPR/Cas9 system (M. A. DeWitt et al., 2016. Selection-free genome editing of the sickle mutation in human adult hematopoietic stem/progenitor cells). They managed to correct a mutation in the cultures of hematopoietic stem cells of patients with an efficiency of 32%, and then introduce them into the body of mice with a defective immune response (a standard approach for studying the behavior of human cells in the body of mice). The resulting cells were preserved in the bloodstream of mice for at least 16 weeks, albeit in a small concentration (about 2% of the total number of hematopoietic cells), but this is most likely enough for clinical improvement. Despite the success, the editing of this mutation has not yet passed into the stage of clinical studies.

In 2020, successful therapy of patients with sickle cell anemia was reported (H. Frangoul et al., 2020. CRISPR-Cas9 Gene Editing for Sickle Cell Disease and β-Thalassemia). In this clinical trial, the therapy was based on a different approach. Adult hemoglobin contains two α-globin and two β-globin molecules. Symptoms of patients with sickle cell anemia begin to manifest during the first year of life, when hemoglobin in red blood cells is replaced by an adult form containing a pathogenic replacement. But before that, no symptoms appear in carrier children, because his red blood cells still contain fetal hemoglobin, which consists of two molecules of α-globin and two - γ–globin. In this work, patients with CRISPR/Cas9 were able to activate the fetal hemoglobin gene in erythrocyte progenitor cells. Clinical studies of this approach to the treatment of sickle cell anemia are still ongoing.

But what stops researchers from a simple approach: correcting a single nucleotide in the genome of patients with sickle cell anemia? One of the main reasons is the shortcomings of the CRISPR/Cas9 system.

Fig. 1. Genome editing scheme using the CRISPR/Cas9 system. DNA endonuclease Cas9 binds to the target genome site using a guide RNA (guide RNA) complementary to a 20-nucleotide site next to the PAM sequence (protospacer adjacent motif). Due to this, a double–stranded DNA break is formed, which the cell seeks to eliminate - for this it has a DNA repair system. Rupture repair is possible in two ways. Repair by non–homologous restoration of the ends is a less precise mechanism that often leads to insertions or deletions. It is this mechanism that researchers rely on when they try to disrupt the gene. Repair using homologous recombination is more accurate, but requires the presence of DNA molecules, which can be used to restore the sequence of the damaged site. When a template containing a given sequence is introduced into cells, you can replace a section of DNA or edit it (including correcting one nucleotide). An illustration from an article by D. Ghosh et al., 2019. CRISPR–Cas9 a boon or bane: the bumpy road ahead to cancer therapeutics.

The repair of the double-stranded break that Cas9 endonuclease forms is based on two DNA repair pathways. Non-homologous connection of the ends (Fig.1, left) – a less precise method that leads to small insertions (inserts of DNA fragments) or deletions (removal of fragments). It is such changes in the genome that are desirable if the purpose of editing is to correct the malfunction of some part of the DNA. For example, in clinical studies of CRISPR therapy for sickle cell anemia, the target DNA site was located in the gene of transcription factor BCL11A. Disruption of BCL11A leads to increased expression of the fetal hemoglobin gene. With this approach, the fact of the gene malfunction itself is important, and not a "letter-by-letter" correction. To correct the replacement of one nucleotide, for example in the hemoglobin B gene, the researchers hoped for a second repair pathway – homologous recombination. In addition to Cas9 and guide RNA, a DNA template with a normal version of the hemoglobin gene is injected into cells. The mutation correction was indeed successful, but in 33% of cases, the authors observed unnecessary deletions and insertions. Which of these two mechanisms a cell will use to repair a double-stranded break depends on the type of cells. In addition, homologous recombination works only in certain phases of the cell cycle, and the two types of repair compete with each other. Most often, a non-homologous connection of the ends wins. In addition, in response to double-stranded DNA breaks, the TP53 gene is activated, which can lead to cell apoptosis.

Genomic editing strategies that allow avoiding double–stranded DNA breaks and their repair began to be developed quite quickly, and in 2016 a new approach to correcting mutations was published - base editing (Fig. 2).

Fig. 2. Methods of editing DNA bases. a is the mechanism of CBE (cytosine base editor, editing of cytosine bases).
Above: CBE consists of a defective Cas9 enzyme (Cas9n nickase, shown in blue) that makes a single-stranded break next to the target DNA sequence. The enzyme cytidine deaminase (orange) and an inhibitor of uracylglycosylase (purple) are attached to Cas9n using a linker. The target site (protospacer) and the PAM sequence are present in genomic DNA (genomic DNA). Bottom: after binding guide RNA (guide RNA, gRNA) with genomic DNA and the formation of a "bubble" cytidine deaminase deaminates cytosine, turning it into uracil, and Cas9 makes a single-stranded break of the DNA chain associated with gRNA. Usually, the insertion of uracil into DNA is corrected by the enzyme uracylglycosylase, but the presence of a peptide inhibiting this enzyme stops this process. During DNA replication or repair, adenine stands opposite uracil on another DNA chain. b – AVE mechanism (adenine base editor, editing of adenine bases).
Above: ABE consists of Cas9n (blue), to which the enzyme adenosine deaminase (orange) is attached using a linker. Bottom: after binding gRNA to genomic DNA and forming a "bubble", adenosine deaminase deaminates adenine, turning it into inosine, and Cas9 makes a single-stranded break of the DNA chain associated with gRNA. During DNA replication or repair, guanine stands opposite inosine on another DNA chain.
Figure from the review by E. M. Porto et al., 2020. Base editing: advances and therapeutic opportunities.

In base editing systems, the Cas9 enzyme is also present, however, in a defective form (which is called a nicase), which makes only one cut of the DNA chain. Such a single-stranded incision marks the nucleotide that needs to be replaced. There is also a guide RNA that directs Cas9 to the target DNA site. But the most important difference is the enzymes that modify the nitrogenous bases on DNA (see Figure 2).

The first system is a method of editing cytosine bases, which replaces complementary pairs of C•G with T•A. It contains the enzyme cytidine deaminase, which deaminates cytosine, converting it into uracil. Normally, excisional base repair mechanisms react to uracil in DNA, and the enzyme uracil glycosylase cuts uracil from DNA. To preserve uracil, the system developers added a phage peptide that inhibits uracil glycosylase.

The second system is a method of editing adenine bases, which replaces T•A pairs with C•G. The system also contains an enzyme that modifies adenine, converting it into inosine. Interestingly, there is no enzyme in nature that could make such a modification in DNA. Therefore, the researchers obtained such an enzyme using artificial evolution based on the enzyme TadA Escherichia coli. Repair systems ignore inosine in DNA, so no additional elements were added to the system. After the conversion of cytosine to uracil or adenine to inosine during DNA repair or replication, adenine or guanine stands in front of them, respectively.

Both methods have seemed attractive from the very beginning to correct single-nucleotide substitutions associated with human diseases. The first publication on CBE describes the successful correction of a mutation in variant IV of the apolipoprotein E (APOE4) gene associated with the development of Alzheimer's disease, as well as a mutation in the TP53 gene associated with certain types of cancer (A. Komor et al., 2016. Programmable editing of a target base in genomic DNA without double-stranded DNA cleavage). In the first publication about ABE, the authors corrected a mutation in the HFE gene that leads to hereditary hemochromatosis, and also modeled mutations in the promoters of the HBG1 and HBG2 genes that ensure the preservation of fetal hemoglobin, which means they can potentially be used to treat sickle cell anemia (N. Gaudelli et al., 2017. Programmable base editing of A•T to G•C in genomic DNA without DNA cleavage).

In an article published recently in the journal Nature, American researchers led by David David R. Liu, the founder of base editing, was able to correct the very mutation in the β–globin gene that leads to the formation of an abnormal form of hemoglobin HbS and sickle cell anemia. True, it is impossible to correct the GTG (Val) codon directly to GAG (Glu) using ABE, but it can be replaced with a naturally occurring non-pathogenic variant of GCG (encoding alanine, Ala) (Hb-Makassar, named after the city of Makassar in Indonesia, where this variant was first discovered).

To correct mutations, the authors used the latest generation of ABE developed by them earlier, which finds target sequences more efficiently and replaces T•A pairs with C•G. Either hybrid protein ABE (Cas9-nicase and adenosine deaminase) together with guide RNA or mRNA encoding ABE and guide RNA were injected into hematopoietic cells of donors with sickle cell anemia. The introduction of protein and RNA resulted in an 80% correction of the mutation, whereas for the mRNA construct the efficiency was 44± 5.9%. However, the second approach (introducing only mRNA) provided fewer unwanted insertions and deletions.

The edited cells were differentiated into erythroid progenitors that express proteins of the β-globin locus. At the same time, five times less defective hemoglobin was found in the edited cells than in the control (Fig. 3, c). The editing also did not affect the possibility of cell differentiation, and also improved the resistance of young erythrocytes (reticulocytes) to hypoxia: they were less willing to turn into sickle cells (Fig. 3, d, e). The effectiveness of this process depended on which method of introducing the system was used, but the authors do not discuss the reasons for such differences.

Fig. 3. Editing of hematopoietic cells of donors with sickle cell anemia. a is a section of the HBB gene to be edited.
The nucleotide A (blue color) is to be replaced by G. This will lead to the replacement of abnormal HBBS hemoglobin with a non-pathogenic form of HHBG (Makassar variant). The PAM sequence is highlighted in purple, the sequence complementary to the target (Protospacer) is highlighted in green. Brown and light green highlighted potential "silent" mutations (see Silent mutation) when replacing adenine with guanine. b is the editing level of the target (A7) and potential adenines (A9, A12 – "silent" mutations, A16 – missense mutation) 6 days after the introduction of the constructs.
Unedited – unedited adenines, mRNA-edited – edited using mRNA constructs, RNP-edited – edited using a hybrid protein introduced together with guide RNA. c is the proportion of β-globin proteins (β-like globins, that is, proteins expressed at the β-globin locus) in reticulocytes: yellow – hemoglobin γ, red – hemoglobin δ, blue – hemoglobin βG (non-pathogenic Hb-Makassar), gray – hemoglobin βS (pathogenic variant).
d – images of reticulocytes from edited and unedited hematopoietic cells of a donor with sickle cell anemia in hypoxia.
Pay attention to the irregular shape of the unedited reticulocytes. The length of the scale segment is 50 microns. e – counting sickle-shaped reticulocytes (sickled cells).
f – estimates of potential non-target mutations proposed by the Cas-OFFinder program (red circle) and the CIRCLE-seq method (yellow circle).
The green color indicates non-target mutations confirmed by sequencing. g – locations of detected non–target mutations: brown – promoter, purple – intron, blue - exon (all mutations are silent), green – intergenic regions, red – 5' UTR (untranslated region, untranslated region), orange – 3'-UTR, yellow – TTS (DNA region of 1000 or less base pairs up to the terminator).
h is the number of confirmed non–target mutations (validated off-target sites) in the target sequence, distributed by the number of genome reads.
A drawing from the discussed article in Nature.

Nevertheless, the authors investigated in detail the possibility of non-targeted mutations. First, they conducted a bioinformatic analysis, with the help of which they found 140 DNA sites containing the PAM sequence and a sequence complementary to the target with the exception of 1-3 bases. Secondly, the authors used an in vitro method for estimating the number of non-target CIRCLE-seq mutations. A total of 725 potential sites were discovered, 697 of which were confirmed by genome sequencing. Only 54 of them corresponded to the replacement of T•A with C•G. When using a combination of ABE and RNA, non-target mutations were less common, probably due to the fact that the complex is active only for a short time (Fig. 3, f–h).

Finally, modified hematopoietic cells were injected into immunodeficient mice and observed how they differentiated into erythrocyte precursors. After 16 weeks, the bone marrow of mice consisted of 70% human cells, while they behaved the same way as control, unmodified cells. In the edited precursors of erythrocytes, the level of defective β-globin decreased by more than two times.

Human erythrocytes have a short life span in the body of mice. To study how the edited precursors of erythrocytes will behave and how this will affect the health of animals, the authors transplanted them to Townes mice (Fig. 4). These mice have their own α- and β-globin genes replaced with human ones, while there is a mutation that causes sickle cell anemia. Mice with two HBBS alleles develop the corresponding symptoms. Mice were injected with modified cells and blood samples were taken after 6, 10, 14 and 16 weeks to assess how functional the modified cells were. At week 10, 90% of the donor cells took root in all mice. The amount of corrected β-globin in erythrocytes was 75-82% throughout the experiment. Such erythrocytes lived longer, since at week 16, modification was observed only in 44% of edited hematopoietic cells.

Fig. 4. Experiment on editing of erythrocyte precursors (CD45.2) Townes mice have sickle cell anemia models. In mice with homozygous HBBS/S mutation (and sickle cell anemia), erythrocyte precursors were isolated from the bone marrow, ABE (ABE8e-NHCH) was injected into this culture by electroporation, and then cells were transplanted into CD45.1 C57Bl/6 mice, in which the bone marrow was previously destroyed. In addition, unedited cells of heterozygous HBVA/S mice were transplanted. After 16 weeks, the results of transplantation were evaluated in groups of mice with edited HBBS/S genotype (Edited HBBS/S), unedited HBBS/S genotype (unedited HBBS/S), HBBA/S genotype compared with the control without transplantation. A drawing from the discussed article in Nature.

A general blood test showed that mice that received unedited cells showed all the signs of sickle cell anemia: not only elongated, but also fragmented red blood cells, as well as a decrease in the level of red blood cells and leukocytes. In mice with edited cells, these parameters were similar to those of healthy mice.

Editing a mutation that causes sickle cell anemia is not the first attempt to use ABE to correct pathogenic mutations. At the beginning of 2021, a group of American researchers led by the same David Liu used a specially developed ABE to correct a mutation in the lamin A gene. This mutation causes Hutchinson–Guilford syndrome, a childhood form of progeria (accelerated aging) (L. W. Koblan et al., 2021. In vivo base editing rescues Hutchinson–Gilford progeria syndrome in mice). It was possible to correct it both in fibroblasts of children with progeria and in a mouse model of this disease. Just one injection of an adeno-associated virus expressing ABE increased the life of mice from 215 to 510 days.

In May 2021, another article was published describing how a mutation was introduced using ABE that disrupts the PCSK9 gene (proprotein convertase of subtilisin-kexin type 9). This gene is mainly expressed in the liver and regulates the receptors of low–density lipoproteins, one of the main carriers of cholesterol in the blood. The PCSK9 protein binds to receptors, causing cholesterol degradation. If there are too few of these receptors, the metabolism of low-density lipoproteins is disrupted and their level in the blood increases. Therefore, a decrease in the activity of this gene should lead to a decrease in cholesterol levels, especially in hereditary forms of hypercholesterolemia. The use of ABE gives a great advantage when working with liver cells. Since hepatocytes are renewed every 200-300 days, double breaks in the DNA in these cells are repaired by non-homologous reconnection of the ends, that is, a more mutagenic pathway. Mutations introduced using ABE lead to disruption of PCSK9 mRNA splicing. In mouse hepatocytes, the editing efficiency was 84 ± 4.6%.

Fig. 5. Directed disruption of the PCSK9 gene in mice. a is the target sequence of guide RNA in the PCSK9 gene in mice (Mus musculus), as well as macaques and humans (Macaca fascicularis/Homo sapiens).
Two target nucleotides are highlighted in bold (T•A will be replaced by C•G), the PAM sequence is indicated in blue, the exon 1 of the PCSK9 gene is in red, and the intron is in black. b – experimental scheme: adenovirus-associated vector (AAV) or lipid particles (LNP) containing ABE were injected into mice.
On the second day, a second dose of the vector (re-dose) was administered. Editing with the introduction of lipid particles was evaluated on day 18, and with the introduction of adenovirus-associated vectors - on day 43. c is the percentage of editing when using different ABE variants.
Purple color – editing the target base, gray color – inserts and deletions (Indels). d is the level of Pcsk9 protein in the blood plasma of mice (gray – mice that did not receive a vector (untreated), purple – mice that received ABE using adenoviruses (AAV, ABEmax).
e is the level of low–density lipoproteins in the blood plasma of mice.
f – level of antibodies (measured by ELISA analysis) to Cas9 and TadA proteins after administration of AAV.
Blood plasma of mice injected with TadA or ready-made antibodies to Cas9 was used as a control. A drawing from the article under discussion in Nature Biotechnology.

Further, the authors used adenovirus-associated vectors to deliver various ABE variants to the liver cells of mice, where after 6 weeks the maximum editing efficiency was 60 ± 18% (Fig. 5). At the same time, the level of low-density lipoproteins decreased by almost five times.

Adeno-associated vectors remain for a long time in non-dividing cells, for example, hepatocytes. This may contribute to the appearance of additional undesirable mutations (in addition to the target) due to the work of ABE. Therefore, the authors tested the delivery of ABE using mRNAs enclosed in a lipid envelope (Fig. 5, b). In this case, editing was observed in 86.9 ± 1.9% of liver cells. As for additional mutations, the authors did not find a significant increase in the transition of A to G and inserts and deletions.

Inspired by the result, the researchers decided to conduct an experiment on crab-eating macaques – one of the closest preclinical models to humans. The monkeys were injected with lipid particles that contained the guide RNA necessary to disrupt the synthesis of PCSK9, as well as the mRNA encoding the most successful ABE variant. The drug was administered once or every two weeks. Monkeys tolerated the introduction of RNA well, blood tests differed slightly from normal ones.

After 29 days of the experiment, the animals were killed and the level of editing of liver cells was evaluated. Even under conditions of double administration of a high dose of the drug, the mutation was observed only in 24.14 ± 1.52% of liver cells. This is significantly lower than in mice. This is a very important result that shows that it is important to use suitable animal models. The level of low-density lipoproteins decreased by 19%.

In both publications, the authors note that the use of ABE in general does not lead to a significant number of undesirable changes in the genome. This is good news in terms of the potential use of base editing in gene therapy. During our lifetime, we accumulate a large number of mutations in cells, therefore, most likely, a small number of additional genome changes will not lead to clinical consequences. The fact that ABES, at least, do not increase the activity of the TP53 gene gives hope that these mutations will not lead to malignant changes and other consequences.

However, there is still a lot of research separating us from the use of base editing in gene therapy. First, it is necessary to develop methods that will reduce the number of insertions and deletions in the cells into which ABE falls. This applies both to the method of introducing the system (in the form of mRNA or in the form of a complex of protein and guide RNA), and to choosing a suitable system in the case of rapidly renewing (for example, bone marrow) or slowly renewing tissues (for example, liver tissues). Secondly, it is important to increase the efficiency of editing, which is still a problem for CBE. In addition, both systems, in addition to DNA, are capable of modifying ribonucleotides in RNA, which can affect protein synthesis in target cells.

In any case, editing the bases is a new and inspiring method on which the medicine of the future can be based.

Sources:

1) Gregory A. Newby et al., Base editing of haematopoietic stem cells rescues sickle cell disease in mice // Nature. 2021.
2) Tanja Rothgangl et al., In vivo adenine base editing of PCSK9 in macaques reduces LDL cholesterol levels // Nature Biotechnology. 2021.

Portal "Eternal youth" http://vechnayamolodost.ru