Introduction

ATP binding cassette (ABC) transporters are trans-membrane ubiquitous proteins, that translocate natural substrates across plasma membranes. In humans, there are 49 genes coding for ABC transporters, arranged in eight subfamilies extensively studied because at least 11 of 49 genes are known to cause severe inherited diseases1. The ATP-binding cassette, sub-family A member 12 (ABCA12) gene was discovered from cDNA of human placenta2. ABCA12 is 207 kb long with fifty-three exons and two very long introns at its beginning (26.5 kb and 47.3 kb, respectively, Fig. 1a). ABCA12 is a keratinocyte transmembrane transporter that binds and hydrolyzes ATP to transport lipids in the lamellar granules3. This activity is required to form the extracellular lipid barrier in the outermost layer of the skin, the stratum corneum of the epidermis3. The lipid barrier is composed of three lipid classes (cholesterol, free fatty acids, and ceramides) and acts as a primary barrier between the body and the environment to prevent excessive water loss and to avoid penetration of pathogens4. ABCA12 has also a role in immunity: in macrophages it regulates the cellular cholesterol metabolism via an LXRb-dependent post-transcriptional mechanism5.

Figure 1
figure 1

Positive selection at ABCA12. (a) The ABCA12 gene has is 55 exons and two long introns at its beginning. The black rectangle indicates the 70 kb region surrounding rs10180970 considered in this project. (b) rs10180970 is the most differentiated variant between Africans and non-Africans as shown by the absolute difference of the derived allele frequency (ΔDAF), however, other variants seem to be also extremely differentiated. (c) The Cross Population Extended Haplotype Homozogysity statistic (XP-EHH), measured between pairs of continental populations, shows a signal of positive selection in non-Africans over 35 kb downstream rs10180970, especially in East-Asians. (d) The signal is confirmed by the Integrated Haplotype Score (iHS) within continental populations. In panels b–d the dashed line indicates the genomic position of rs10180970.

Full size image

UV-radiation has a major effect on skin and keratinocytes and it is one of the most studied environmental stressors of the epidermal homeostasis. In keratinocytes, UV-radiation induces mutagenesis, apoptosis, proliferation, and metabolic changes. UV-radiation has a major role in determining skin pigmentation6,7, but it also can have a detrimental effect on the lipid barrier reducing the intercellular lipid cohesion8. Earlier studies suggest that UVB radiation down-regulates epidermal ABCA12 gene expression9 and in general UV light influences other ABC transporters activity in lymphocytes10.

The always growing availability of ancient DNA information allows us to directly reconstruct the history of genetic mutations and understand if consequences of past adaptations are relevant for contemporary humans11. Past events of positive selection can be tested as detection of a major shift in allele frequencies12,13,14.

In a genome-wide scan for positive selection in contemporary humans, we identified a signal of positive selection in Europeans and Asians at rs10180970 C/T, located in the second intron of ABCA1215. In the study, we demonstrated that the derived allele T is highly frequent in Asians and Europeans, compared to Africans, where the ancestral C variant prevails. This evidence, together with the identification of patterns typical of positive selection in the genomic region of rs10180970 let us speculate that ancestor of current Europeans and Asians should have had an advantage in carrying the T allele and that it would be interesting to determine the functional consequences and the trajectory of the T allele through time.

Here we present results from investigating the genomic region surrounding rs10180970. We confirm that this region underwent positive selection and investigate one possible selected phenotype. We have extended the set of individuals used to investigate genetic variation and natural selection at rs10180970, including also DNA sequences from ancient human samples. We confirm the signal of positive selection at rs10180970, investigate linkage with genetic variants in its proximity and the effect of the two alleles on ABCA12 gene expression.

Results

Natural selection signal at the ABCA12 gene extends 10 kb downstream rs10180970

We identified rs10180970 as a possible candidate for positive selection in Europeans and Asians during a genome-wide scan for positive selection conducted on populations from Phase I 1000 Genomes Project15. In this study, we considered a 70 kb region surrounding rs10180970 including the first two exons, the first and most of the second introns, and the transcription starting site of ABCA12 (Fig. 1a). Compared to Phase I, the 1000 Genomes Phase III dataset16 includes 1,412 more samples belonging to 12 more populations (Supplementary Table S1) and in the region under study includes 1,612 variants for which the ancestral state is known and that we used to calculate statistics suggestive of population differentiation and positive selection.

Between pairs of continental populations, we evaluated the absolute difference of the derived allele frequency (ΔDAF) and the Cross Population Extended Haplotype Homozygosity (XP-EHH17). ΔDAF approaches 1 when the derived allele prevails in one population and is almost absent in the other. According to ΔDAF, rs10180970 is still the most differentiated variant in comparisons of Africans and non-Africans in 70 kb surrounding rs10180970 (Fig. 1b). In Phase III the T/T genotype frequency in the combined sample of Europeans and Asians is significantly higher than Africans (Table 1, Fisher exact test p-value < 2.2e-16). Similarly, in the Complete Genomics dataset (CG)18, the T/T genotype is significantly more frequent among Asians and Europeans compared to Africans. This is true both when considering samples partially overlapping with Phase I, and samples unique to CG (Table 1, Fisher exact test p-value < 2.2e-16), suggesting that the observed pattern of higher derived allele frequencies in Europeans and Asians compared to Africans was not due to the specific set of sample used but is reproducible. XP-EHH detects recent positive selection highlighting selective sweeps in which the selected allele has approached or achieved fixation in one population but remains polymorphic in the overall population. In the 1000 Genomes Phase III dataset, we observe a stretch of negative XP-EHH outside the expected variance range (i.e., -1,1) in pairs of Africans with non-Africans. The signal is located approximately at 10 kb downstream rs10180970, suggesting recent positive selection in non-Africans in this region (Fig. 1c), and it is particularly strong in East-Asians. Within continental populations we calculated the Integrated Haplotype Score (iHS19). iHS detects the classic signal of strong directional selection through the identification of unusually long haplotypes with low diversity surrounding core SNPs. As for XP-EHH, also for iHS we observe peaks of unusually high iHS 10 kb downstream rs10180970 in non-African populations and especially in East-Asians (Fig. 1d).

Table 1 Genotypes frequencies at rs10180970 in Phase III 1000 Genome and Complete Genomics datasets both full and non overlapping with 1000 Genomes Phase I (unique).
Full size table

Altogether these results show that while rs10180970 is the most differentiated variant in the ABCA12 gene, haplotypes downstream rs10180970 differ between Africans and non-Africans. Haplotypes carrying the T allele of rs10180970 are very common in Europeans and Asians and tend to be longer and less diverse in these two populations compared to the others, suggesting that this region might have been affected by recent positive selection in these two populations.

As linkage between variants in a long haplotype can be responsible for spurious signals, we computed linkage disequilibrium (LD) in continental populations between rs10180970 and 2,034 genetic variants in 70 kb surrounding rs10180970. For all populations but Africans, we observe LD within 10 kb downstream rs10180970 (Fig. 2a). In East-Asians and Europeans LD extends over 35 kb downstream rs10180970, in a region inclusive of the first intron and the first exon. Within this block, LD is high in East-Asians (r2 around 0.75) and moderate in Europeans (r2 around 0.50). The genomic region in high LD contains functional features such as chromatin regulation (H3K4Me1 and H3K27Ac), enhancers and a promoter located in the proximity of the first intron and includes SNPs with ΔDAF > 0.70 (Fig. 2b). Therefore we conclude that other SNPs are in high LD with rs10180970 in East-Asians and part of a long haplotype that is located in a genomic region with regulatory features.

Figure 2
figure 2

Linkage disequilibrium between rs10180970 and variants within 70 kb. (a) Vertical dotted line indicates the genomic position of rs10180970. Circles represent other variants at their genomic position on the x-axis and the r2 measures the non-random association of the alleles of each variant with rs10180970 on the y-axis. Crosses distinguish pairs in which the genetic variants have ΔDAF > 0.70 as shown in Fig. 1. (b) Screen-shot form Genome Browser showing chromatin modification and state segmentation for several cell types in the genomic region under investigation.

Full size image

No significant association between ABCA12 expression and genotype at rs10180970

Having confirmed the signal of natural selection at rs10180970 we next asked if the two alleles have a different effect on gene expression. Furthermore, in the hypothesis that another variant in linkage with rs10180970 can be responsible for the functional changes, we decided to include also rs2970968 in the analysis. Besides being in linkage disequilibrium with rs10180970 in East-Asians and having high ΔDAF like other variants, rs2970968 has two further features that make it an interesting variant. First, we predicted the functional consequences of the variants with high ΔDAF and in linkage disequilibrium with rs10180970 in East-Asians using Funseq20 (Supplementary Table S2). We confirmed that rs2970968 (leftmost dark red cross in Fig. 2a) is located in a region predicted to be an enhancer21 in lymphoblastoid, HeLa and HUVEC cell lines and that it is located in transcription factor binding peak of several transcription factors, although not in transcription factor motifs within peaks (Supplementary Fig. S1). One of the transcription factors is implicated in mammalian skin tumors (ATF222,23) and three others are regulated by UVB radiation (STAT3, GATA3, and POU2F2)24,25,26. Secondly, rs2970968 is the closest SNP to a region that has shown to be essential for the functioning of the promoter of ABCA1227 (Supplementary Fig. S1).

We investigated ABCA12 expression in relation to the three genotypes of rs10180970 (C/C, C/T, T/T) and rs2970968 (A/A, A/G, G/G). Human skin would be the ideal tissue to test allele-specific gene expression, however sampling technique is invasive. It is instead possible to quantify ABCA12 expression in total mRNA extracted from pulled hairs without the need for invasive skin biopsies28. Because ABCA12 expression is influenced by sun exposure29, to discriminate between the genetic and the environmental effects, we performed the experiment in a group of individuals of different geographical origin but all resident in the same city. We determined the genotypes at rs10180970 and rs2970968 through DNA sequencing and measured ABCA12 expression through quantitative PCR of reverse transcribed total mRNA extracted from pulled hair roots. We considered as reference the average ABCA12 gene expression of genotypes of the ancestral alleles (C/C for rs10180970, and AA for rs2970968), which are also the allele prevalent in Africans, and calculated the fold change of the other observed genotypes. After quality controls, we had data on both genotype and ABCA12 expression for 36 and 39 samples at rs10180970 and rs2970968, respectively. As shown in Fig. 3, there is no significant correlation between ABCA12 gene expression and genotypes at both loci (One-way non-parametric ANOVA p-values > 0.9 for both sites), suggesting no effect of rs10180970 or rs2970968 genotypes on ABCA12 expression in hair roots. Therefore, we conclude that the effect of the selection at these two loci on gene expression is not obvious when individuals are exposed to similar UV radiation.

Figure 3
figure 3

ABCA12 expression in relation to the three genotypes of (a) rs10180970 and (b) rs2970968. Fold change refers to average ABCA12 gene expression of ancestral genotypes, which are also the genotypes prevalent in Africans (C/C for rs10180970, and AA for rs2970968).

Full size image

While it is not possible to model the gene effect in mouse because there is no homologous region in mouse, we found that ABCA12 is well expressed in HaCaT cells, a spontaneously transformed aneuploid immortal keratinocyte cell line from adult human skin30. We determined by Sanger sequencing that HaCaT cells are heterozygous at rs10180970 and therefore can be used for allele-specific expression assessment at this locus. As a preliminary work, we silenced ABCA12 in HaCatT cells using small interfering RNAs and confirmed that ABCA12 expression is reduced after UVB radiation (20 mJ and 30 mJ), with downregulation remaining sustained up to 18 hours after treatment (Fig. 4).

Figure 4
figure 4

ABCA12 expression in HaCaT cells and UVB irradiation (a) ABCA12 is expressed in HaCaT cells and can be silenced using small interfering RNA. (b) ABCA12 expression is reduced after Ultraviolet B (UVB) radiation and cells do not recover even after 18 hours. T6 and T18 indicate 6 and 18 hours after UVB radiation; mJ = milliJoule.

Full size image

The T variant of rs10180970 is specific to H. sapiens and rapidly increased in frequency in non-Africans

We determined the pattern of genetic variation at rs10180970 in time and space in 88 ancient and 2,764 contemporary samples ranging the last 80,000 years with the aim to understand when the T allele first appeared and how its frequency evolved through time. Among contemporary samples, 41 were collected for this study and are from 9 geographically diverse populations (Supplementary Table S1). Their genotype at rs10180970 was determined by Sanger sequencing a 600 bp region containing the variant. For the rest of the contemporary samples genotypes at rs10180970 were publicly available from the 1000 Genomes Phase III16 and a study on West Africans31. For ancient samples, we used publicly available raw sequencing data (Supplementary Table S3) to perform variant calling. Beside H. sapiens, ancient samples include also one Denisova sample and two Neanderthals.

The allele frequency of the derived allele T (fT) through time was determined grouping individuals in bins of 2000 years (Fig. 5). Some pools are made of very few individuals (n = 1 in extreme cases). Individuals were subdivided according to species and within H. sapiens in Africans and non-Africans. In our data, the T allele appeared for the first time around 45 kya in a H. sapiens sample from Russia32 and rapidly fT reached 0.9 in non-Africans remaining stable until the present day.

Figure 5
figure 5

Prevalence of the T allele of rs10180970 in time and space. (a) T allele trajectory in the last 80,000 years in 88 ancient and 2,764 contemporary samples. The T allele is observed for the first time 45,000 years ago in non-Africans and rapidly rose in frequency. (b) T allele frequency is very low among African populations with the exception of North-Africans (especially Egyptians and Ethiopians) and African American. (c) T allele was probably recently introgressed in African American, as admixture analysis show a shared genetic component of African American with Europeans and East-Asians in the region under investigation. Vertical bars represent single individuals and colors reflect membership to admixture clusters in the most likely hypothesis of five clusters.

Full size image

Of the 16 African ancient samples available to date, only three have reliable data in the rs10180970 region (Supplementary Table S3). The most ancient African data is from Ethiopia 4.5 kya33, while the other two are from 1.3-2 kya34. All ancient Africans have genotype C/C. In contemporary Africans fT is very low but not completely absent (0.09), while in North-Africans fT is on average 0.4 and T is the major allele in Egyptians, within North-Africans (Fig. 5b). Admixture analysis shows that the Africans and North-Africans with higher fT, in fact, have quite high European and South Asian components probably reflecting recent admixture (Fig. 5c).

Both Neanderthals and the one Denisova sample available are homozygous for the ancestral C allele, suggesting that the T allele might have been specific to H. sapiens, as far as we know about the other two species. In support of this hypothesis, two recent publications find that there is a very little probability of admixture in the region surrounding rs10180970 between H. sapiens and Neanderthals35 (Supplementary Fig. S2) and H. sapiens and Denisova36.

We conclude that the T variant of rs10180970 is specific to H. sapiens and rapidly increased in frequency in non-Africans, while the low prevalence of T in contemporary Africans might be due to recent admixture with Europeans and South-Asians.

Discussion

Humans are exposed to environmental agents through the naked skin, and in fact, strong natural selection has been detected in a number of skin-related phenotypic traits such as hair and eye pigmentation6,37,38, and hair thickness39,40. The lipid barrier is the most external layer of the skin and it acts as an interface with the environment, providing protection against UV, dehydration, and pathogens. The formation of lipid barrier is a complex process and mutations in genes involved in this process result in skin diseases. Nevertheless, to date, to our knowledge, no studies investigated natural selection of genes that contribute to build the lipid barrier such as ABCA12.

The genetic variant rs10180970 located on chromosome 2 in the ABCA12 gene, is predicted to be under positive selection in Europeans and Asians, as we discovered in a genome-wide scan based on population differentiation15. Here we investigated the genomic region containing rs10180970 to ask whether the variant is truly under positive selection and investigated the functional consequences of the selection event.

We confirmed the signal of selection in Europeans and Asians and discovered high linkage disequilibrium in a 35 kb region downstream rs10180970 (Fig. 2). In the 35 kb region the signals of positive selection are as strong or stronger than those initially found at rs10180970 (Fig. 1) and include the promoter of the gene (Fig. 2). By using a larger dataset of DNA sequences and a more fine population subdivision compared to the one used to discover the signal, we observe that the signal is stronger in East-Asians compared to South-Asians, possibly because South-Asians exposure to UV radiation is more similar to that of Africans. Despite confirming the positive selection signal we could not observe obvious allele-specific effect of rs10180970 on ABCA12 gene expression (Fig. 3) in a sample of individuals exposed to the same level of UV radiation. Finally, we traced the history through time of the allele under selection using publicly available ancient human sequences and established that the mutation is specific to H. sapiens (Fig. 5, Supplementary Fig S2). While our findings strongly corroborate the signal of selection they also suggest that a different functional approach should be used to decipher the functional consequences of the event of selection in terms of gene expression.

We suggest that further work is required to fully decipher the effect of selection on ABCA12 and identified the keratinocyte-derived HaCaT cells as a model system for in vitro experiments. We determined that ABCA12 is expressed and can be silenced in the HaCaT cells, and more importantly HaCaT cells are heterozygous at rs10180970, therefore ideal for investigating allele-specific effects. On the other hand, because of the possibility that other sites in linkage with rs10180970 are involved, it would be probably worth while to consider sequence information at other sites as well. More in general, overall indication for future research implicate the study of the non-coding regions of ABCA12, like sequence/haplotype-specific enhancer assay, even if the dimensions of the gene can be a limitation.

There are two main findings from our work. First, we reconstructed the demographic history of the T allele using the currently available information to discover that it is specific to H. sapiens and it was apparently very frequent in humans that migrated out of Africa prior to 50 kya. The second main finding is that while rs10180970 might act as a sentinel for the signal of selection, the signal extends over 35 kb and includes the first intron, the first two exons and the transcription starting site of ABCA12. Linkage disequilibrium in this region is high in East-Asians and Europeans, there are several regions predicted to be enhancers and two regions predicted to be sites of chromatin modifications (methylation and acetylation), thus possibly influencing gene expression.

Our study endorses and refines previous studies. Similar to ABCA12, other members of the ABC family have also been identified in previous scans for positive selection41. The genomic region containing ABCA12 was identified as one of the most differentiated between continental populations42, and in fact ABCA12 hosts one of the top ten SNPs in the genome, useful to discriminate continental populations in clustering analyses43. ABCA12 was identified as one of the top candidate genes in a study investigating local adaptation in Asians44 and stands out in a selection scan performed with LD-based statistics on HapMap Phase II data45. However, none of these studies provides an indication of the exact genomic location of the signal within the gene. In this study, we corroborate the signal of selection at rs10180970 that was localized in our previous study15, and provide initial evidence that will help understand why the T allele of rs10180970 was favored in the ancestors of current Asians, in particular, East-Asians, and Europeans. Our work also shows that with the exception of few cases (e.g.46,47) it is surprisingly difficult to identify the selected allele and selected phenotype, even with abundant sequence data and functional tests.

Methods

Samples and genotyping

Forty samples were collected from healthy fully informed, consenting, adult volunteers from 9 countries (Supplementary Table S1). Saliva samples were collected using the Oragene DNA (OG-500) kit. Genomic DNA was phenol–chloroform extracted from saliva48. All experiments were carried out in accordance with relevant guidelines and regulations. The study was approved by the ethics committee of the University of Naples Federico II (protocol number 106/15) and conducted in accordance with the relevant guidelines and regulations. Written informed consent was obtained from all participants.

Genotypes at rs10180970 and rs2970968 were obtained by Sanger-sequencing a 325 bp and 400 bp regions, respectively, of the ABCA12 gene where the variants are located. The fragments were amplified from genomic DNA by PCR in a final volume of 50 μl using oligos designed with Primer349,50 and the following conditions: 1 cycle 95 °C for 5′; 35 cycles at 95 °C for 1′, 62 °C for 1′, 72 °C for 45″; 1 cycle at 72 °C for 10′. PCR products were purified using PCR purification kit from QIAGEN.

All other contemporary samples in the study are part of public collections and genotype data at rs10180970 was available in form of vcf files16,18,31. For ancient samples we screened publicly available sequence reads data of 400 samples spanning 80,000–200 years before present32,51,52,53,54,55,56,57,58,59,60,61,62,63,64, however filtering for coverage >2 in the region of interest resulted in 89 samples listed in Supplementary Table S3. Variant calling in ancient samples was performed with samtools/bcftools65.

Derived allele frequencies, linkage disequilibrium, admixture, introgression, natural selection analyses

Derived allele frequencies were computed using vcftools66 using the information on derived allele from the 1000 Genomes Project16. Linkage disequilibrium (LD) between rs10180970 and any other SNP in ±2 Mb was computed with PLINK1.967. Clustering analysis in a region of 2 Mb surrounding rs10180970 including 55,683 Single Nucleotide Polymorphisms (SNPs) was done with ADMIXTURE68. ADMIXTURE was run under the hypotheses of 2 to 10 clusters with cross-validation computation.

Data for checking if the region of rs10180970 is among the ones putatively introgressed form Neanderthal were downloaded from http://genetics.med.harvard.edu/reichlab/Reich_Lab/Datasets_-_Neandertal_Introgression.html.

Integrated haplotype score (iHS) and Cross-Population Extended Haplotype Homozygosity (XP-EHH) were calculated with Hapbin69. Parameters for iHS calculation were–minmaf 0.01, –cutoff 0.01. iHS and XP-EHH log-ratio were normalized to have zero mean and unit variance using the standard normalization implemented in Hapbin that groups variants in bins of 2% frequency.

RNA extraction and processing from hair follicles

Hair follicles (5 to 7 bulbs) were grasped as near to the scalp as possible without damaging hair roots. Total RNA was extracted from bulbs on the day of collection, or following storage at 4 °C for 1–5 days in RNA Later and placed into a 1,5 ml microcentrifuge tube containing 1 ml of Trizol Reagent (Fisher Molecular Biology). RNA extraction was then conducted according to manufacturer’s instructions. 500 ng RNA were reverse-transcribed using SuperScript® VILOTM MasterMix (Invitrogen). qPCR was performed on technical triplicates per each sample using Light Cycler 480® II (Roche), on 2 μl of previously diluted cDNA (1:5) template using LightCycler® 480 SYBR Green I Master Mix (Roche), according to the manufacturer instructions. The thermal profile consisted of 1 cycle at 95 °C for 10′; 40 cycles at 95 °C for 10″, 60 °C for 10″, 72 °C for 10″. Fold changes were calculated relative to Hypoxanthine Phospho Ribosyl Transferase (HPRT1) expression as in70.

HaCat cells culturing, genotyping, silencing and UV radiation

HaCat cells were cultured in 75 cm2 tissue culture flasks in Dulbecco’s Modified Eagle’s Medium, containing 10% fetal bovine serum and 1% penicillin and streptomycin. Cells were maintained in a humidified atmosphere (5% CO2) at 37 °C. Genomic DNA was extracted from HaCat cells71 and genotype at rs10180970 was determined by Sanger sequencing.

For silencing experiments, HaCaT cells were transfected with a ABCA12-targeting siRna (Sigma Aldrich). Transfection was performed with Oligofectamine following the manufacturer’s protocol. Cells were collected 72 hours after transfection. Silencing efficiency was evaluated by Real Time Quantitative PCR (RT-qPCR) following total RNA extraction using RNeasy Mini Kit-QIAGEN and 1 μg and reverse-transcription using SuperScript® VILOTM MasterMix (Invitrogen).

For UVB radiation HaCaT cells were plated in 60 mm plates and grown to 70–90% confluence. Subsequently, cells were irradiated with 20 and 30 mJ/cm2 using three lamps (Philips Ultraviolet 8 TL 20 W/01 RS lamps; Philips, Eindhoven, Netherlands) generating UVB light in the range of 290–320 nm with an emission peak at 312 nm. The intensity of UVB irradiation was measured using a phototherapy radiometer (International Light, Newburyport, MA) of UVB. After irradiation cells were incubated for 0, 6 and 18 hours and ABCA12 expression was assessed by Real Time PCR (RT-PCR). Fold changes were calculated relative to Ribosomal Protein Large P0 (RPLP0) expression as in70,72.