Gene Conversion

DT40^Cre1, which displays increased Ig gene conversion due to a v-myb transgene and contains a tamoxifen-inducible Cre recombinase, has been described previously (Arakawa et al. 2001 (link)). DT40^Cre1AID^–/– was generated by the targeted disruption of both AID alleles of DT40^Cre1 (Arakawa et al. 2002 (link)). AID^R was derived from DT40^Cre1AID^–/– after stable integration of a floxed AID-IRES-GFP bicistronic cassette, in which both AID and GFP are expressed from the same β-actin promoter. AID^RψV^– was derived from AID^R by transfection of pψVDel1-25 (see Figure 1A). Stable transfectants that had integrated the construct into the rearranged light chain locus were then identified by locus-specific PCR. Targeted integration of pψVDel1-25 results in the deletion of the entire ψV gene loci starting 0.4 kb upstream of ψV25 and ending 1 bp downstream of ψV1. AID^RψV^partial was produced in a similar way as was AID^RψV^–, by transfection of pψVDel3-25, which, upon targeted integration, leads to a partial deletion of the ψV loci starting 0.4 kb upstream of ψV25 and ending 1 bp downstream of ψV3. Cell culture and electroporation were performed as previously described (Arakawa et al. 2002 (link)). XRCC3^–/– was derived from DT40^Cre1 by deleting amino acids 72–170 of the XRCC3 gene following transfection of XRCC3 knockout constructs. Clones that underwent targeted integration were initially identified by long-range PCR, and the XRCC3 deletion was then confirmed by Southern blot analysis.

To investigate if accelerated evolution in TFBSs was driven by positive selection, we used the INSIGHT model to infer positive selection on the seven accelerated, non-overlapping TFBS groups in the human lineage^{50 (link)–52 (link)}. We obtained INSIGHT2, a highly efficient implementation of the INSIGHT model, from https://github.com/CshlSiepelLab/FitCons2. Then, we applied INSIGHT2 to each TFBS group under accelerated evolution and the collection of all TFBSs from ENCODE. INSIGHT2 provided D_p and SE[D_p], that is, the expected number of adaptive substitutions per kilobase and its standard error, as well as ρ and SE[ρ] which quantified the fraction of sites under selection within functional elements and its standard error. We performed the Wald test to examine if D_p was significantly different from 0 for each TFBS group. Under the null hypothesis of D_p = 0, we assumed that the z-statistic, $\frac{D_p}{\text{SE}\left[D_p\right]}$ , asymptotically followed a 50:50 mixture of a point probability mass at 0 and a half standard normal distribution⁸⁸ . We conducted comparisons of ρ among the seven accelerated TFBS groups and the collection of all TFBSs from ENCODE (Supplementary Table 2). To identify the role of GC-biased gene conversion in the accelerated evolution of the seven TFBS groups, we used the phastBias model to infer gene conversion disparity B in the lineage where accelerated evolution occurred, identical to the best-fit lineage found in model comparison (Fig. 7).

The following fly lines were used: hypomorphic hh^bar3 (FlyBase ID FBal0031487), hh null (hh^AC/TM3, Bloomington #1749), nubHackFLP (described below), nubbin AC-62 (nubbin-Gal4, Bloomington #25754), hh[KO;FRT]/TM3 (ref. ^{34 (link)}, kindly provided by J.P Vincent), w¹¹¹⁸ and FRT82B kindly provided by C. Klämbt, FRT82B Rps3 (bearing Gal80, Gal80 not being used in this study as a tool) kindly provided by S. Luschnig. The nubHackFLP flies used also beared additional markers without influence on the results and of no relevance to the present study. Injection of pRIV^white carrying the wild-type hh sequence or hh variants coding for R/A and R/E variant yielded hh[KO;hh]/TM3, hh[KO;hh^R/A]/TM3, and hh[KO;hh^R/E]/TM3 (this lab). For clonal analysis of homozygous alleles of all forms, a wing-blade-specific FLP source on chromosome II was generated. To this end, nubbin-Gal4 was chosen as the target for CRISPR-mediated gene conversion by using the pHackGal4FLP (described earlier) derivative of the Hack-Gal4 Vector pBPGUw-Hack-QF2 (Addgene #80276), following the method of ref. ^{80 (link)}. pHackGal4FLP was injected in a Cas9; nubbin AC-6-bearing background. Subsequent screening of the F1 offspring for positive convertants, as revealed by the presence of 3P3-DSRED, yielded 12 integration events from 500 injected embryos. 3P3-DSRED was then floxed out and isogenic lines were established. We refer to this chromosome as nubFLP. We used the following genotype for clonal analysis of the hh rescue constructs in the eye and wing disks (Fig. 3): yw eyFLP3.5/yw; nubFLP/+; FRT82B (Gal80) Rps3/FRT82B, hh[KO;hh^transgene or no hh]. This enabled us to study clones in eyes and wings in the same animals. The resulting fly eye phenotypes were captured with a Nikon SMZ25 microscope and quantified by using Nikon F-package software version 4.5.01. w¹¹¹⁸ flies or heterozygous flies of the indicated genotypes served as positive controls. Wings of the adult F1 were collected and mounted in Hoyer’s medium, and total intervein areas and total wing areas were quantified by MoticImage software version 2.0. All fly crosses and maintenance were conducted at 25 °C. For wing quantifications, ten male and ten female wings were analyzed for each data set and ratios between L3-L4 intervein areas and L2-L3 intervein areas were determined.