The major problem with the first generation base editors(BE1) included the formationof undesired products due to the followingtwo reasons: (i) frequent removal of uracil by cellularN-glycosylase (UNG) and (ii) possible occurrence of more than one cytosines (Cs) within the target window of 4-8 bases, which would allow base editingof non-target cytosines also, inaddition to the target cytosine. Theenzyme UNG works during Base Excision Repair (BER) and therefore, will treat transitional edited base pair U:G (derived from C:G) as DNA damage, so that U of U:G base pair will be excised,and U:G will not be able to produce T:A, leading to a failure of desiredconversion of C:G into T:A. Keeping this in view and in order to increase in vivo editing efficiency,second generation base editors (BE2)were developed, which carried auracil glycosylase inhibitor (UGI) fused with dCas9, so that the enzymeUNGwill be inhibited and will not beable to excise U fromthe U:G base pair, whichwill be converted to T:A during DNA replication.
The editing efficiency of these 2nd generation base editors (BE2)was improved (reaching amaximum of ~20 and formationof indels significantly reduced (<0.1%) overthat obtained in CRTISPR-mediated genome editing. The second problem of the occurrence ofmore than one Cs in the editing window was partly resolved by reducing the size of editing window from 4-8 basepairs to 1 or 2 base pairs (see later). The next stage of improvement of baseeditors was achieved by converting dCas9to a nickase throughreplacement of either amino acid aspartate (D) by alanine (A) at position 10 (D10A; also describedas nCas9), or replacement of amino acid histidine (H) by alanine at position 840 (H840A). nCas9 and H840A both produce nicks in opposite strands, and have been suitably utilized insingle base gene editing14(Ran et al., 2013). For instanceD10A mutant of Cas9 retains a domain that generates a single strand DNA nick inthe non-target strand instead ofcreating double strand breaks at the desired site; this wouldsimulate mismatch repair, sothat a unmodified opposite DNA strand wouldmimic a DNA strand undergoing synthesis, where the strand containing the editedbase is used as a template (C®U; Fig.
4), taking U as T. Therefore, BE3 had the followingthree components: (i) an AID/APOBEC1 deaminase, that was fused to a Streptococcus pyogenes nuclease deficient nickase Cas9nCas9n(D10A), and (iii) a UGI that was linked to Cas9n through a 4 amino acids linker. Theimportance of UGI in base editing was demonstrated by showing that theUGI-deleted BE3 (BE3-?UGI ) was less competent in base editing compared to original BE3, and produced not only lower frequency of desired C®T editing, but also produced a higher frequency of unwanted indels. Anumber of improved BE3 variants were also developed (Table 2), which resultedin much more efficient conversion of the G:U intermediate to desired A:U and A:T products4,11 (Komor et al.
, 2016, 2017). Another problem associated with BE1 and BE2 was the occurrence of more than one Cs within thebase editing window,so that the cytosine deaminase will convert even a non-targeted C into U. Thisproblem was overcome by the development of BE3 with SpCas9 (NGG), where eventhe non-NGG PAM sequence could be used for base editing (see later). It was also shownthat addition of another copy of UGI to BE3 further reduced the frequency ofindels, so that BEs with more than one UGI were developed and were described as4th generation base editors, the BE4, which were found to be more efficient (Wang et al.
, 2017). BE4 or SaBE4 were furtherimproved by adding Gam to thecassette, so that the use of BE4-Gam resulted in a further 1.5 to 2.0 fold decrease in the indelfrequency (Table 1).