On the genetic basis of tail-loss evolution in humans and apes

Abstract

The loss of the tail is among the most notable anatomical changes to have occurred along the evolutionary lineage leading to humans and to the 'anthropomorphous apes'^1-3, with a proposed role in contributing to human bipedalism^4-6. Yet, the genetic mechanism that facilitated tail-loss evolution in hominoids remains unknown. Here we present evidence that an individual insertion of an Alu element in the genome of the hominoid ancestor may have contributed to tail-loss evolution. We demonstrate that this Alu element-inserted into an intron of the TBXT gene^7-9-pairs with a neighbouring ancestral Alu element encoded in the reverse genomic orientation and leads to a hominoid-specific alternative splicing event. To study the effect of this splicing event, we generated multiple mouse models that express both full-length and exon-skipped isoforms of Tbxt, mimicking the expression pattern of its hominoid orthologue TBXT. Mice expressing both Tbxt isoforms exhibit a complete absence of the tail or a shortened tail depending on the relative abundance of Tbxt isoforms expressed at the embryonic tail bud. These results support the notion that the exon-skipped transcript is sufficient to induce a tail-loss phenotype. Moreover, mice expressing the exon-skipped Tbxt isoform develop neural tube defects, a condition that affects approximately 1 in 1,000 neonates in humans¹⁰. Thus, tail-loss evolution may have been associated with an adaptive cost of the potential for neural tube defects, which continue to affect human health today.

Fig. 1. Evolution of tail loss in hominoids.

a, Tail phenotypes across the primate phylogenetic tree. Ma, millions of years ago. b, UCSC Genome browser view of the conservation score through multi-species alignment at the TBXT locus across primate genomes. Exon numbering of human TBXT follows a conventional order across species without including the 5′ untranslated region exon. The hominoid-specific AluY element is highlighted in red. LINE, long interspersed nuclear element; LTR, long terminal repeat; SINE, short interspersed nuclear element. c, Schematic of the proposed mechanism of tail-loss evolution in hominoids. Primate images in a and c were created using BioRender (https://biorender.com).

Fig. 2. Both AluY and AluSx1 are required for inducing alternative splicing of TBXT.

a, CRISPR-generated homozygous knockouts of the AluY element in TBXT intron 6 (top, TBXT^{ΔAluY/ΔAluY}) and AluSx1 element in intron 5 (bottom, TBXT^{ΔAluSx1/ΔAluSx1}) in human ES cells. b, RT–PCR results of TBXT transcripts isolated from differentiated human ES cell of wild-type, TBXT^{ΔAluY/ΔAluY} and TBXT^{ΔAluSx1/ΔAluSx1} genotypes. Each mutant line included two independent clones. All RT–PCR results were performed in technical duplicates. c, A schematic of inferred Alu interactions and the corresponding TBXT transcripts, which indicate that an AluY–AluSx1 interaction leads to the TBXT^Δexon6 transcript. The TBXT^Δexon6–7 transcript may stem from an AluSx1–AluSq2 interaction.

Fig. 3. The TBXT^Δexon6 isoform is sufficient to induce tail-loss phenotype.

a, CRISPR design for generating the Tbxt^Δexon6/+ heterozygous mouse model. b, Schematic of TBXT transcripts in human and mouse models. Tbxt^Δexon6/+ mouse mimics TBXT gene expression products in humans. c, Sanger sequencing of the RT–PCR product confirmed that deleting exon 6 in mouse Tbxt leads to correct splicing by fusing exons 5 and 7. d, A representative Tbxt^Δexon6/+ founder mouse (day 1) exhibiting a no-tail phenotype. Two additional founder mice are shown in Extended Data Fig. 5. e, Tbxt^Δexon6/+ mice exhibit heterogeneous tail phenotypes, varying from no tail to long tails. cv, caudal vertebrae; sv, sacral vertebrae; WT, wild type; arrowheads highlight differences in tail phenotypes.

Fig. 4. Introducing inverted intronic sequence pairs induces short-tail phenotypes in mouse models.

a, Schematic of the mouse Tbxt gene structure with the inserted human AluSx1–AluY pair (Tbxt^insASAY). The engineering of the Tbxt^insASAY model involved a two-step strategy specified in Extended Data Fig. 7a (Methods). b, Gene structure of the Tbxt^insRCS2 model with an insertion of a 220 bp RCS from intron 6 to intron 5 of Tbxt (Methods). c,d, Tail length of Tbxt^insASAY mice (c) and Tbxt^insRCS2 mice (d) across age, grouped by sex and genotypes. Tbxt^+/+ is the wild type. Data in c and d are presented as the mean ± s.d. of tail length (mm) in the corresponding group. Each mouse group included 4–11 mice from multiple litters, with dots indicating individual data points of the group. e, Tailbud-expressed Tbxt transcripts detected by RT–PCR using E10.5 mouse embryos across genotypes from Tbxt^insASAY (left) or Tbxt^insRCS2 (right) intercrossing experiments. RT–PCR results are presented as biological duplicates, with consistent results obtained from more independent embryos across genotypes. f, Representative tail phenotypes across mouse lines, including wild type, Tbxt^{insASAY/insASAY}, Tbxt^{insRCS2/insRCS2} and Tbxt^{insRCS2/Δexon6}. Each included both male (M) and female (F) mice. g, Summary schematic of the correspondence between the relative abundance of Tbxt isoforms in mice of different genotypes and their observed tail phenotypes.

Extended Data Fig. 1. Comparative genomics analyses of hominoid-specific variants in the genes related to tail development.

a, Workflow of the comparative genomics analyses (Methods). b, Summary of all detected hominoid-specific variants with respect to the outgroup species.

Extended Data Fig. 2. RNA structure prediction.

a, Predicted RNA secondary structure of the TBXT exon5-to-exon7 sequence using the RNAfold algorithm of the ViennaRNA package. The paired AluY-AluSx1 region was highlighted. b, Mountain plot of the RNA secondary structure prediction, showing the ‘height’ in predicted secondary structure across the nucleotide positions. Height was computed as the number of base pairs enclosing the base at a given position. Overall, the AluSx1 and AluY regions were predicted to form helices with high probability (low entropy).

Extended Data Fig. 3. Analyses of TBXT isoforms.

a, In vitro differentiation of human and mouse ESCs for inducing TBXT/Tbxt expression. Human and mouse ESCs differentiation assay was adapted from Xi et al. (2017) and Pour et al. (2019), respectively. Schematic adapted from icons created by Marcel Tisch via bioicons.com. b, Quantitative RT-PCR (RT-qPCR) of TBXT and MIXL1 expression during hESC differentiation, indicating correct induction of mesodermal gene expression program. c, Quantitative RT-PCR of Tbxt expression during mESC differentiation. Data in b and c were presented as mean +/− standard deviation of the relative gene expression levels. Sample number n = 3 represents RT-qPCR results from three biologically independent RNA samples, with each data point averaged from 3 technical replicates in quantitative PCR. d, RT-PCR of TBXT/Tbxt transcripts in human and mouse differentiated ESCs, highlighting the Δexon6 splicing isoform unique to human. RT-PCR results were presented as biological duplicates. e, Protein sequence alignment of TBXT-exon6 region in the representative mammals. All presented animals have tails except human and chimpanzee. f, The exon 6-coded peptide of Tbxt protein overlaps with large fractions of two transcription regulation domains. TA, transcription activation; TR, transcription repression. Functional domain annotation of mouse Tbxt protein was adapted from Kispert et al. (1995).

Extended Data Fig. 4. Validation of Alu-deletion hESC clones and the expressed TBXT isoforms.

a, PCR validation of the hESC clones with deletions of AluY or AluSx1 in TBXT. PCR validation for each sample were performed in pairs, each amplifying both AluSx1 locus (Sx1) and the AluY locus (Y) using primers that bind the two flanking sequences of the targeted region, respectively. Each genotype included two independent clones of AluY deletion or AluSx1 deletion, corresponding to the two clones presented in Fig. 2b. b, Sanger sequencing of the TBXT^Δexon6 and TBXT^Δexon6-7 transcripts detected in Fig. 2b. The sequencing results were aligned to the TBXT full length mRNA sequence.

Extended Data Fig. 5. Tbxt^Δexon6/+ founder mice generated through zygotic CRISPR/Cas9 targeting approach.

a, Schematic of zygotic injection of CRISPR/Cas9 reactions. b, Two more Tbxt^Δexon6/+ founder mice (in addition to the one shown in Fig. 3d) indicating an absence or reduced form of the tail. c, Sanger sequencing of the exon 6-deleted allele isolated from the genomic DNA of Tbxt^Δexon6/+ founder mice. Founder 1 had an unexpected insertion of 23 base pairs at the CRISPR cutting site in the original intron 5 of Tbxt. Both founder 2 and 3 had the exact fusion between the two CRISPR cutting sites in introns 5 and 6. d–e, Capture-seq analyses at the Tbxt locus of founder mice did not detect off-target mutations. A zoomed-in view of the Capture-seq results at the Tbxt locus highlights the CRISPR-mediated exon 6 deletion (e). Capture-seq baits were generated from bacterial artificial chromosomes (RP24-88H3 and RP23-159G7). The shallow-covered regions are typically repeat sequences in the mouse genome and are consistent across samples. Control DNAs were obtained from wild-type or Tbxt^{Δexon6/Δexon6} mESCs, and the heterozygous sample came from a 1:1 mixture of genomic DNA from wild-type and Tbxt^{Δexon6/Δexon6} mESCs.

Extended Data Fig. 6. Engineering of inverted sequence pairs in mouse Tbxt induces alternative splicing.

a–b, Schematics of mouse Tbxt gene structure with inserted human AluSx1-AluY pair (a, Tbxt-insASAY) or a designed intronic reverse complementary sequence of 297 bp (b, Tbxt-insRCS). The designed RCS insertion has the same length as AluY in human TBXT. In both designs, a two-step experimental procedure was adapted by first integrating the target elements with a selection cassette of puromycin-resistance and truncated thymidine kinase (puro-ΔTK) gene into the intron of mouse Tbxt, followed by removal of the selection cassette through transiently expressing Cre recombinase (Methods). c, Tbxt transcripts detected through RT-PCR using differentiated mouse ESC lines across wild-type (left), homozygous Tbxt-insASAY (Tbxt^{insASAY/insASAY}, middle), and homozygous Tbxt-insRCS (Tbxt^{insRCS/insRCS}, right) genotypes. RT-PCR results were presented as biological duplicates for each genotype. d, mESC injection into diploid or tetraploid blastocyst for generating Tbxt^{insASAY/insASAY} and Tbxt^{insRCS/insRCS} mouse models.

Extended Data Fig. 7. Validation of Tbxt^{insASAY/insASAY} and Tbxt^{insRCS2/insRCS2} homozygous mice.

a-b, Capture-seq reads mapped to mouse reference genome mm10 at the full Tbxt locus (a) and a zoom-in view of Tbxt gene region (b) in Tbxt^{insASAY/insASAY} and Tbxt^{insRCS2/insRCS2} homozygous mice. Black ticks under each coverage track indicate the detected SNVs referring to the mm10 genome (a). The black bars in (b) indicate the detection of reads supporting an inversion. As expected, Tbxt^{insRCS2/insRCS2} samples incorporated an intronic sequence insertion, thus resembling an inversion event at the insertion site (b) due to forced mapping of the sequencing reads to reference genome. c, Sanger sequencing of Tbxt^{insRCS2/insRCS2} genomic sequence confirmed the inserted sequence and the exact insertion site. The inserted sequence constitutes a 220 bp sequence from Tbxt-intron 6 (mm10 chr17: 8439335-8439554). d, Sanger sequencing of RT-PCR results using total RNA extracted from tailbud of Tbxt^{insRCS2/insRCS2} embryo at stage E10.5. The results correspond to Fig. 4e.

Extended Data Fig. 8. Exon6 deletion of Tbxt may lead to neural tube defects in mouse.

a, Analyzing E11.5 Tbxt^{Δexon6/Δexon6} mouse embryos obtained through intercrossing Tbxt^Δexon6/+ mice. Tbxt^{Δexon6/Δexon6} embryos either developed neural tube closure defects (middle) that died at birth or arrested at approximately stage E9 during development (right). Red and black dashed lines mark the embryonic tail regions and limb buds, respectively. Green arrowheads indicate malformed spinal cord regions. b-c, Both Tbxt^{Δexon6/Δexon6} (b) and Tbxt^Δexon6/+ (c) neonatal mice may present neural tube closure defects during embryonic development. The presented embryos were the only two cases found in this study that died after birth with neural tube closure defects.

Extended Data Fig. 9. RNA-seq analysis of Tbxt target genes in differentiated mESC lines across genotypes.

The analyzed mESC lines include wild-type, Tbxt^{insASAY/insASAY}, Tbxt^Δexon6/+, and Tbxt^{Δexon6/Δexon6} genotypes, each in duplicates. a, Scatter plot of the samples using the first two principal component (PC) coordinates in principal component analysis. b, Heatmap of Tbxt-target genes that were differentially expressed across the analyzed samples. Tbxt-target gene list was obtained from Tbxt ChIP-seq results using in vitro differentiated mESCs. Functionally characterized Tbxt-target genes were labeled on the y-axis of the heatmap. c–e, Volcano plots of differentially expressed (DE) genes comparing mutant mESCs with the wild-type mESC. DE genes were identified using DESeq2 (version 1.40.2) through its default two-sided Wald test and a cutoff of log2 fold expression change >0.5 and multiple test-adjusted p value (p.adj) <0.05. For each plot, DE Tbxt-target genes were highlighted in red, and the top DE genes among this group were labeled.

Extended Data Fig. 10. A model for tail-loss evolution in the early hominoids.

The AluY insertion in TBXT may have marked a key genetic event that contributed to tail-loss evolution in the hominoid common ancestor. Additional genetic changes – pre-existing in the ancestral genome or occurring after AluY insertion – may have also acted to promote or stabilize the no-tail phenotype in the early hominoids.