Abstract

   Actively retrotransposing primate-specific Alu repeats display
   insertion-deletion (InDel) polymorphism through their insertion at new
   loci. In the global datasets, Indian populations remain
   under-represented and so do their Alu InDels. Here, we report the
   genomic landscape of Alu InDels from the recently released 1021 Indian
   Genomes (IndiGen) (available at
   [46]https://clingen.igib.res.in/indigen). We identified 9239
   polymorphic Alu insertions that include private (3831), rare (3974) and
   common (1434) insertions with an average of 770 insertions per
   individual. We achieved an 89% PCR validation of the predicted
   genotypes in 94 samples tested. About 60% of identified InDels are
   unique to IndiGen when compared to other global datasets; 23% of sites
   were shared with both SGDP and HGSVC; among these, 58% (1289 sites)
   were common polymorphisms in IndiGen. The insertions not only show a
   bias for genic regions, with a preference for introns but also for the
   associated genes showing enrichment for processes like cell
   morphogenesis and neurogenesis (P-value < 0.05). Approximately, 60% of
   InDels mapped to genes present in the OMIM database. Finally, we show
   that 558 InDels can serve as ancestry informative markers to segregate
   global populations. This study provides a valuable resource for
   baseline Alu InDels that would be useful in population genomics.

INTRODUCTION

   Indian populations with their complex demographic history are extremely
   diverse. They contain thousands of endogamous sub-populations from
   different ethnic and linguistic lineages, with varying levels of
   admixture as well as social structure. There are four major linguistic
   lineages: Indo-European (IE), Dravidian (DR), Tibeto-Burman (TB) and
   Austro-Asiatic (AA) ([47]1). Distinct geographical and climatic clines
   further contribute to this population diversity. So far, estimates of
   genetic diversity within India and its relatedness with global
   populations have been studied extensively using single nucleotide
   polymorphisms (SNPs) ([48]1,[49]2). Primate-specific Alu elements,
   present in more than a million copies in the human genome, also serve
   as informative markers for understanding the genetic diversity of
   populations ([50]3–5). Since SNPs can occur due to replication errors,
   not all SNPs are identical by descent. Moreover, each Alu insertion
   creates a structural feature of approximately 300 bp, and therefore,
   they may be inherently more likely to have a practical consequence than
   an SNP. The younger subfamilies of Alu (AluY) are still
   retro-transpositionally active ([51]6). AluYa5 is currently the most
   active Alu subfamily in the human lineage, followed by AluYb8, and many
   others including the four newly identified subfamilies termed as
   AluYb7a3, AluYb8b1, AluYa4a1 and AluYb10 ([52]7,[53]8). Active
   transposition of these into newer sites contributes to Alu
   insertion-deletion polymorphism (InDels) in the genome ([54]9). Once
   retrotransposed, these are stable and define a biallelic locus based on
   their presence/absence at specific locations in the human genome
   ([55]10). The absence of the Alu in the loci of interest is considered
   as the ancestral state and is regarded as the deletion allele ([56]10).
   Polymorphic Alu elements are therefore identical by descent and this
   feature makes them more informative compared to other genetic markers
   such as SNPs. Thus, Alu InDels are of enormous utility as ancestry
   informative markers in population genomics and association studies
   ([57]11).

   Notwithstanding this, identification of Alu insertion-deletion
   polymorphism has been tenuous as their size, repetitive nature and
   abundance make them challenging to discover and annotate even by high
   throughput genomic methods ([58]12,[59]13). Recently, next-generation
   genome sequencing with a higher depth of coverage in diverse
   populations has started yielding these polymorphic markers from
   different populations. One of the prominent resources for population
   genetic studies is the 1000 Genomes project that includes approximately
   489 individuals related to India (Phase 3 release, last accessed 15
   September 2021) ([60]14). These samples have been sequenced with higher
   coverage in the Human Genome Structural Variation Consortium (HGSVC), a
   resource specific to structural variants ([61]15). The South Asians or
   Indian populations represented in these datasets are mostly admixed
   populations and do not represent the entire genetic spectrum of India.
   Another attempt by the Simon Genome Diversity Project (SGDP) used 296
   individuals representing 49 South Asians of which only 21 samples were
   from Indians ([62]16).

   Recently, the IndiGen project has provided the whole genome sequences
   of over 1021 individuals from different geographical locations of India
   ([63]17). The high depth of coverage of these genomes allowed us to
   explore the Alu insertion-deletion (InDels) landscape in these
   populations. We compared IndiGen Alu InDels with those reported in some
   of the publicly available datasets (HGSVC and SGDP) and also studied
   their patterns within genomes. We report a total of 9239 polymorphic
   Alu insertions in Indian genomes out of which 60% are unique to
   IndiGen. These include 1434 common insertions with frequency ≥5%, and
   the remaining fraction includes private and rare insertions (with
   frequency <5%). We could experimentally validate 84 of the 94 predicted
   genotypes. The polymorphic insertions show significant bias for genic
   regions and are significantly enriched in cell morphogenesis and
   neurogenesis processes. Approximately 60% also map to loci implicated
   in Mendelian diseases. A set of 558 Alu insertions are ancestry
   informative and can distinguish world populations based on their
   genetic affinities. This study provides an enormous resource for
   genome-wide Alu InDels in the Indian population that would be useful in
   population genomics, disease associations as well as functional
   genomics studies.

MATERIALS AND METHODS

Study population and datasets

IndiGen

   Raw BAM files were obtained from the whole genome sequencing of 1021
   young, healthy, unrelated Indian individuals sequenced as a part of the
   IndiGen study ([64]17). The samples were sequenced on Illumina NovaSeq
   6000 platform (Illumina Inc. San Diego, CA, USA) and data were
   generated as 150 × 2 bp paired-end reads with ∼25–30× coverage and were
   mapped to the human genome build GRCh38/hg38.

Global datasets

   Alu insertions were also retrieved from 296 samples of Simons Genome
   Diversity Project (SGDP) that houses data on seven major world
   populations which we obtained on 25 June 2021 on request from the lead
   author ([65]16) and 3202 samples of the Human Genome Structural
   Variation Consortium (HGSVC last accessed on 15 September 2021)
   ([66]15). The latter includes high coverage genotypes of the 2504
   samples from the 1000 Genomes Project Phase 3 release. These were used
   for comparison with our IndiGen dataset ([67]15). Both these datasets
   had comparable coverage (∼30×) as IndiGen.

Pipeline for identification of Alu insertions

   The MELT (Mobile Element Locator Tool version 2.1.5, last accessed on
   15 September 2021) ([68]18) was used to detect the polymorphic Alu
   InDels as it has been earlier shown to outperform other tools in terms
   of accuracy, sensitivity, scalability and runtime ([69]19) and had also
   been used in global Genome diversity projects (1000 Genomes Phase 3,
   HGSVC and SGDP) to identify Alu insertions. BAM files were used in the
   MELT- SPLIT pipeline for the identification of polymorphic Alus and
   private insertions. The identified sites were annotated with the prefix
   Alu_IndiGen_Alu_ with a bash script. The chromosome-wise count, the
   distribution of sites within genes (in exonic, intronic, UTRs, upstream
   (up to 5 kb before a gene start site), and downstream regions (up to 5
   kb after a gene end site) were analyzed and plotted using R ([70]20).

Quality checking and filtering

   To obtain a high-quality data set for downstream analyses and to avoid
   false positives and missing genotypes, the raw MELT calls were filtered
   stringently. First, the sites with no calls (ac0 flag by MELT) and
   those with >10% missing genotype calls were removed. Then, sites with a
   PASS flag by MELT, a flanking target-site duplication (TSD) defined by
   a MELT ASSESS score of five and in Hardy–Weinberg equilibrium (HWE) in
   the population were retained. HWE analysis was carried out with PLINK
   v1.9 ([71]21). Sites that were either in (i) low complexity regions,
   (ii) not genotyped in >25% samples (s25), (iii) did not have enough
   supporting discordant mapped reads, (iv) without a genotyped allele
   (Allele count 0 filter [ac0] which were removed in the first filtering
   step), (v) biased reads only from one end i.e. 3′ or 5′ of the
   predicted insertion site (rSD) and (vi) split discordant filter (hDP)
   are not marked with the PASS flag and were removed ([72]Supplementary
   Figure S1).

Analysis of the identified insertions

   Variant Effect Predictor (VEP version 104; GRCh38/hg38) ([73]22) was
   used to annotate the identified Alu insertions for their location in
   the genome. For selecting the consequence of a variant insertion in a
   gene, the results were filtered based on ‘one selected consequence per
   variant’ criteria in VEP. MELT annotations were used for assessing
   subfamily distribution. The numbers of polymorphic Alu InDels and their
   density for chromosomal regions split into 10 MB contiguous bins, i.e.
   percentage of Alu insertions occupying 10 MB regions of each
   chromosome, were calculated with customized R scripts. The correlation
   analyses of genic Alu insertion density with GC content, intron
   density, gene density for 1 MB chromosomal regions, and the number of
   insertions with intron lengths and gene lengths were performed using R
   scripts. The annotations of GC content, genes, and introns were
   downloaded from the UCSC Table browser Gencodev36 human genome build
   GRCh38/hg38.

Experimental validation of identified insertions

   We carried out experiment validation of a set of polymorphic Alu
   insertions chosen based on their frequency group. For each of these
   sites, we selected six different IndiGen samples, two each of
   homozygous insertion (Ins/Ins), deletion (Del/Del) and heterozygous
   (Ins/Del) genotypes that were identified from the genome analysis.
   ([74]Supplementary Table S1). The sample sets would therefore vary
   based on the locus studied. We designed primers flanking the site of
   insertions such that an amplified product without the Alu insertion
   would give a product of ∼200–300 bp and with an Alu insertion that of
   ∼500–600 bp (Figure [75]2B). Primers were designed using NCBI primer
   blast from ∼200 bp upstream and downstream DNA sequence of each target
   insertion site ([76]Supplementary Table S2). Polymerase Chain Reactions
   were performed using oligos synthesized by Eurofins with ∼20 ng genomic
   DNA in a 10 μl volume reaction using Taq DNA polymerase (GeNeI, Cat no.
   MME23L). The reaction was carried out on Veriti™ 96-Well Thermal
   CyclerGreen (Cat no: 4375786). The cycling conditions were: 3 min at
   95°C, {30 s at 95°C, 30 s at 55°C (except for InDel_15446 Ta for which
   was 57°C), 30 s at 72°C}X30 cycles, 3 min at 72°C. Insertion amplicons
   were confirmed using Sanger sequencing. Briefly, PCR products were
   cleaned up using SureExtract PCR/Gel Extraction Kit (Genetix Biotech
   Asia Pvt. Ltd., NP-36107) as per the manufacturer’s protocol before
   Sanger sequencing (ABI 3130/3730) using BigDye Terminator v3.1 (ABI,
   Thermo Scientific, California, USA) chemistry. For Sanger sequencing,
   the products were purified using the PEG purification method
   ([77]https://openwetware.org/wiki/PEG_purification_of_PCR_products),
   and the reactions were set with either forward or the reverse primer.
   The cycling conditions were: 3 minutes at 95°C, (10 s at 95°C, 10 s at
   55°C, 4 min at 60°C) X40 cycles. UCSC Blat was done to confirm the
   position of the sequenced amplicon using the FASTA files generated by
   Chromas 2.6.5 and the presence of the Alu insertion was confirmed using
   rmblast of RepeatMasker v3.0 with default parameters.

Figure 2.

   [78]Figure 2.
   [79]Open in a new tab

   Correlation of Alus with GC content, gene density, gene length, intron
   length and intron density. (A) Polymorphic genic Alu InDels density
   identified in 1021 IndiGenomes. (B) Fixed Alus in the reference human
   genome retrieved from the UCSC genome browser GRCh38/hg38. * marks
   parameters where r values for all chromosomes are significant. For
   correlation with GC content and gene density, only a few chromosomes
   did not pass the significance cut-off as provided in [80]Supplementary
   Table S3. Dotted lines connecting the different points is to show the
   trend across different chromosomes.

Comparison with global datasets

   Novel Alu InDels in the IndiGen samples were discovered through
   comparisons with the HGSVC and SGDP datasets. Many Mobile Element
   Insertions (MEIs) discovered in the two datasets had identical
   positions. To account for the positional differences contributed by
   Target Site Duplication (TSD) length and the respective Alu coordinates
   assigned by different MELT versions, we allowed small windows of
   positional tolerance (up to ±50 bp). The overlap was substantially
   increased for up to ±20 bp, especially with HGSVC data, and hence this
   cut-off was used to compare the positions of Alus in these datasets.
   ([81]Supplementary Figure S2)

Population genetics

   We wanted to ascertain the utility of the common Alu InDel
   polymorphisms for population genomics studies. To identify the minimum
   number of insertions required to differentiate among the populations,
   we carried out PCA analysis with Alu insertions. PCA analysis with
   Plink (v1.07) ([82]21) using the genotype data of the sites shared
   between the IndiGen, SGDP and HGSVC was performed. F[ST] analysis was
   carried out with VCFTOOLS using the Weir-Cockerham estimator. F[ST]
   values for the insertions were calculated across the major ancestral
   groups, i.e. Europeans, East Asians, Africans and South Asians (IndiGen
   included), and PCA analysis was also performed with the top (75%, 50%,
   25% and 10%) differentiating Alu InDels.

Pathway enrichment analysis

   The ToppGene (ToppFun) ([83]https://toppgene.cchmc.org/enrichment.jsp)
   ([84]23) was used to perform molecular function and biological
   processes analysis of the genes with Alu InDels. P < 0.05 was set as
   the threshold value. Pathways and processes that crossed significance
   cut-off of q values FDR B&Y < 0.05 are reported.

RESULTS

Identification of polymorphic Alu insertions in IndiGen

   We identified 22 109 potential Alu insertions from the whole genome
   sequencing data of 1021 individuals using the MELT-SPLIT pipeline.
   After the stringent quality filtering steps (detailed in the Materials
   and Methods section; [85]Supplementary Figure S1), 9239 polymorphic Alu
   insertions were retained with an average of 770 insertions per
   individual. About 90% of the insertions were >250 bp ([86]Supplementary
   Figure S3a) which implies that the majority of them are full-length
   insertions mediated by retrotransposition events. Also, the target site
   duplication (TSD) length distribution of Alu insertions varied from 1
   to 29 bp; however, the mean TSD length was around 15 bp
   ([87]Supplementary Figure S3b). In general, the number of identified
   insertions was observed to be proportional to the size of the
   chromosomes with chromosome 2 having maximum insertions and chromosome
   Y the least (Figure [88]1A). However, the density of the insertions did
   not correlate with the chromosome size. On average, there were ∼27
   polymorphic insertions per 10 MB region of the autosomes. The density
   of Alu insertions was highest in chromosome 4 and lowest in chromosome
   Y (Figure [89]1B).

Figure 1.

   [90]Figure 1.
   [91]Open in a new tab

   Distribution of identified polymorphic Alu InDels in 1021 IndiGenomes
   (A) Number of polymorphic Alu InDels in each chromosome (B) Number of
   polymorphic InDels per 10MB region of a chromosome split into
   contiguous bins. Insertions with a frequency ≥5% are common, <5% are
   rare, and present in one individual in IndiGen data are termed as
   private. (C) Distribution of insertions in AluY subfamily; Inset shows
   the distribution in the major subfamilies AluY, AluS and AluJ. (D)
   Distribution of Alu Insertions within a gene; genic versus intergenic
   region is shown in the inset.

Frequency distribution of Alu InDel polymorphisms

   Insertion frequencies for the Alu inserts varied from as low as 0.04%
   to near fixation (99.90%) in the Indian population. In total, we
   observed 3831 private insertions (present in a single individual), 3974
   insertions that were rare (frequency <5%) and 1434 common insertions
   with frequency ≥5%.

Subfamily distribution of Alu InDels

   Since the most recent subfamily of Alus is retrotranspositionally
   active, we next studied their representation in the polymorphic Alu
   insertions. As anticipated, 99.3% of insertions were of the AluY
   family, followed by 0.6% from AluS and with the minimal representation
   from the oldest family AluJ (0.1%) (Figure [92]1C, inset).
   Approximately 45% of AluY insertions are contributed by the subfamilies
   AluYa5, AluYa4 and AluYb8 (Figure [93]1C and [94]Supplementary Figure
   S4a). AluSz contributed the maximum number of insertions in the AluS
   ([95]Supplementary Figure S4b). Only seven AluJo insertions were
   identified.

Patterns of distribution of InDels in the genome

   Alu repeats have been earlier reported to have a non-random
   distribution ([96]24). We, therefore, wanted to see whether the
   polymorphic Alu insertions also have a preference for specific genomic
   regions. To our surprise, 72.3% (6683) of Alu insertions were observed
   within the coding and regulatory regions (Figure 1D, inset). Among
   these, 79% of the sites were intronic, ∼14% inserted in upstream and
   downstream regions and only 1% were exonic. A minor fraction (<1.5%) of
   the sites also mapped to the 3′UTRs and 5’UTRs (Figure [97]1D). We then
   assessed the correlation of different genomic features with the number
   and density of polymorphic Alu insertions (Figure [98]2A). Overall,
   polymorphic Alu insertions are positively correlated with the intron
   length although the extent of the correlation differs across
   chromosomes ([99]Supplementary Figure S5). There was also a significant
   positive correlation (P-value=2.2 × 10^–16) of Alu insertions with gene
   lengths across all chromosomes. Noteworthy, we observed a significant
   positive association between intron density and genic Alu density
   (P-value=5.2 × 10^–04) but not between gene density and genic Alu
   density (P-value=0.081) ([100]Supplementary Table 3a). Chromosomes 9,
   13–16 and 22 exhibited the highest correlations with GC content
   ([101]Supplementary Table S3b). The Alu insertion density showed a
   significant negative association in the intergenic regions (data not
   shown). We also observed a similar pattern of correlation with the
   whole-genome fixed Alus that were retrieved from the UCSC table browser
   human genome build GRCh38/hg38 (Figure [102]2B). This biased
   distribution of Alu insertion-deletion polymorphism corroborates with
   their overall biased representations that have been reported from the
   first draft of human genome sequencing projects ([103]24–26).

Experimental validation of polymorphic insertions

   About 84 of the 94 (i.e. 89%) MELT predicted genotypes were validated
   through PCR and Sanger sequencing ([104]Supplementary Table S4, see
   Materials and Methods). Since polymorphic Alu insertions are biallelic
   markers, there are three possible genotypes viz, homozygous insertion
   (Ins/Ins), heterozygous (Ins/Del) and homozygous deletions (Del/Del)
   (Figure [105]3A). We could validate all three genotypes for 17 out of
   18 insertions selected, which included 10 common, 4 rare and 3 private
   insertions. However, 1 private insertion could not be validated. A
   representative image of different genotypes of a subset of 12 Alu
   insertions is shown in Figure [106]3B. Details of the represented Alu
   insertions are given in Table [107]1.

Figure 3.

   [108]Figure 3.
   [109]Open in a new tab

   Validation of polymorphic Alu InDels identified in 1021 Indigen samples
   (A) Schematic of validation approach for selected polymorphic loci, PCR
   primers marked with red arrows are designed flanking the site of Alu
   insertions leading to expected amplicons of different sizes with and
   without Alu insertions; the three possible genotypes are also shown (C)
   Representative gel electrophoresis image of the three genotypes:
   Ins/Ins (single amplicon at ∼600 bp), Ins/Del (two amplicons; insertion
   at ∼600 bp and deletion at ∼300 bp) and Del/Del (single band at ∼300
   bp) for loci listed in Table [110]1. The band at ∼850 bp in InDel_15446
   is non-specific.

Table 1.

   Details of polymorphic Alu insertions that are represented in Figure
   [111]3B.
   Expected amplicon size (bp)
   S.No. Gene ID No Insertion With insertion Alu Size (bp) Frequency Group
   1 FRAS1 InDel_5797 298 579 281 Common
   2 LRRK2 InDel_15374 281 558 277 Common
   3 SLC30A9 InDel_5542 254 535 281 Common
   4 SEMA6D InDel_17932 253 533 280 Common
   5 COL4A2 InDel_16968 245 526 281 Common
   6 NBAS_1 InDel_1848 300 581 281 Common
   7 PTPRN2 InDel_10992 254 533 279 Common
   8 XDH InDel_1977 220 501 281 Common
   9 CSMD1 InDel_11019 204 484 280 Rare
   10 ITPR1 InDel_3672 280 561 281 Rare
   11 IL17RD InDel_4112 298 579 281 Rare
   12 HDAC7 InDel_15446 254 534 280 Private
   [112]Open in a new tab

   In all the cases RepeatMasker identified the presence of an AluY
   element in the amplicons as expected ([113]Supplementary Figure S6a-i).
   The two private insertions validated i.e. InDel_15446 and InDel_5507
   were located in intron 6 of the HDAC7 gene and 3′UTR (exon 4/4) of TLR1
   gene, respectively. The presence of Alu Y insertion in the sequenced
   amplicon from positions 88 to 399 bases in InDel_15446 was confirmed
   using Repeat Masker. (Figure [114]3B and [115]Supplementary Figure
   S6g). InDel_5507 showed the presence of AluY element from 115–393 bases
   ([116]Supplementary Figure S6h). Though most of the validated
   insertions were intronic, few were present in the UTRs as well. For
   instance, InDel_4893 is a rare variant present in the 3′UTR (exon 3/3)
   of the PTX3 gene, with the AluY element present from 55 to 343 bases.
   This position of Alu overlaps with an enhancer element in UCSC implying
   the presence of Alu could have an impact on regulation
   ([117]Supplementary Figure S6i). Overall, we could experimentally
   validate 91% of the polymorphic Alu insertions identified in our study.

Comparison with the global datasets

   We compared IndiGen data with HGSVC and SGDP datasets and observed 60%
   (5570) of Alu insertions to be unique to the IndiGen data (novel)
   (Figure [118]4A). Approximately half of the shared polymorphic Alu
   insertions between IndiGen, HGSVC, and SGDP are common polymorphisms
   with minor allele frequency (MAF ≥5%) in all these datasets. The
   remaining sites reported are found in different frequencies across the
   populations compared. For example, 88 sites with a frequency >5% in
   IndiGen have comparatively lower frequencies (rare polymorphisms) in
   the other two global datasets. About 68 common insertions in IndiGen
   are found to be rare in the South Asians of HGSVC. A very few variants
   that were private to IndiGen overlapped with other datasets as well
   (214 with SGDP and 40 of these with HGSVC) and hence could not be
   called private insertions in a global perspective, but we refer to them
   as ‘private to IndiGen’ ([119]Supplementary Table S5 and Figure S7). A
   summary of the three datasets is provided in Table [120]2.

Figure 4.

   [121]Figure 4.
   [122]Open in a new tab

   Comparison of IndiGen with HGSVC and SGDP. (A) Venn diagram
   representing the overlap between polymorphic Alu InDels in IndiGen,
   HGSVC and SGDP. (B and C) Principal component analysis (PCA) plots of
   major world populations depicting clustering of each population (B)
   with 2232 polymorphic Alu InDels shared between IndiGen data, HGSVC and
   SGDP, (C) 554 polymorphic Alu InDels sorted on basis of F[ST] value
   (top 25%). The segregation of the population clusters is as good as
   using all the shared Alu InDels. The proportion of variances for PC1
   and PC2 are shown in brackets.

Table 2.

   Summary statistics from IndiGen, HGSVC and SGDP datasets
   Parameters IndiGen HGSVC SGDP
   Sample size 1021 3202 (687 of South Asian ancestry) 296 (49 of South
   Asian ancestry)
   Coverage 25–30× 30× 30×
   Total QC filtered Alu insertions 9239 9331 11 661
   MAF ≥ 5% (common polymorphisms) 1434 3546 1941
   Average insertion sites per individual 614 1705 835
   [123]Open in a new tab

Utility as Ancestry Informative Markers (AIMs)

   Polymorphic Alu elements have been used as ancestry informative markers
   in population genetic studies ([124]3). We wanted to ascertain the
   utility of the common Alu InDel polymorphisms for population genomics
   studies. PCA analysis using Plink (v1.07) ([125]21) using the genotype
   data of the 24% sites (2232 sites) shared between IndiGen, HGSVC and
   SGDP revealed their proximity to the South Asian populations of the
   latter two global datasets (Figure [126]4B). Only a small percentage of
   IndiGen (71 samples) that were of the Tibeto-Burman ancestry was closer
   to the East Asians as expected (data for the analysis with SNPs are
   unpublished. Similar results were observed with Alus as AIMs, data not
   shown). Further, to identify the minimum number of insertions required
   to differentiate between the populations we carried out PCA analysis
   with Alu insertions, all insertions as well as top 75%, 50%, 25% and
   10% insertions with high F[ST] values. A minimum number of 223 Alu
   insertions could cluster different populations. However, 558 insertions
   i.e. top 25% with high F[ST] showed results as good as all shared
   insertions (Figure [127]4C). About 58.5% of the shared sites in the
   three datasets were observed with a common frequency (MAF ≥ 5%) in
   IndiGen. For each set of Alus sorted based on their F[ST] values, about
   50% of the insertions were found to be of common frequency in IndiGen
   (MAF ≥ 5%) ([128]Supplementary Figure S7 and Table S6).

Functional impact of Alu insertions

   About 6683 (72.3% of total insertions) of the genic insertions mapped
   to 4209 genes, 60% of which are present in genes reported in the OMIM
   database ([129]27) implying the likely importance of these insertions
   in Mendelian diseases ([130]Supplementary Figure S8). Toppfun ([131]23)
   biological pathway analysis with q-value FDR B&Y < 0.05 of all the
   genes revealed significant enrichment of biological processes like cell
   morphogenesis, cell adhesion, nervous system development, axonogenesis
   and synaptic transmission ([132]Supplementary Table S7). Since Alu
   elements have been implicated in neurodevelopment and neurological
   diseases ([133]28), we wanted to see how many genes in our data are
   implicated in neurological diseases. For this, we intersected (Venny)
   our gene list with the NDDVD database that has 289 genes associated
   with 37 different neurodegenerative diseases. We found that 62 out of
   289 genes (i.e. 21%) have Alu insertions in them. 24 of those genes
   have been implicated in Alzheimer’s disease. ([134]Supplementary Figure
   S9a, b).

DISCUSSION

   Polymorphic Alu insertions arise due to recent retrotransposition
   events in the human genome. Alu insertion/deletion polymorphisms have
   been of enormous utility in population genomics studies (as they are
   one of the most informative markers for inferring ancestry), forensic
   applications and disease association studies ([135]3,[136]29,[137]30).
   In this study, we report the genomic landscape of the polymorphic Alu
   insertions in 1021 Indian individuals. There were 9239 polymorphic Alus
   with an average of 770 insertions per individual. Earlier studies have
   reported that the average number of polymorphic Alu insertions per
   individual vary from 1283 ([138]31) to 1574 ([139]32). Most of the
   insertions observed were of full length as shown by their size
   distribution and variable target site duplication length, suggesting
   that they could be transposed by canonical L1 transposase activity
   ([140]33,[141]34).

   We observe that ∼99% of the polymorphic insertions are contributed by
   the most retrotranspositionally active AluY subfamily followed by AluS
   and very few from AluJ. Though older Alu subfamilies AluS and AluJ
   elements have been presumed to be inactive for the past 35 million
   years, there are reports of some of them being active in the human
   genome. For instance, Bennett et al. report four insertions from
   ancient AluS subfamilies two of them were intact Alus and two were
   fragmented copies ([142]31). In another study done by Mills et al.,
   3.3% of total Alu insertions were found to be from the AluS subfamily
   in a comparison of human and chimpanzee retrotransposon insertions
   ([143]35). The Alu elements harbor a large number of regulatory motifs
   and retrotransposition of these elements could provide novel regulatory
   sites ([144]36–39). The insertions predominantly map to the coding and
   regulatory regions compared to intergenic regions. Of these, nearly 79%
   were in the introns and their densities significantly correlated with
   the length of the introns as well as genes (P-value< 2.2 × 10^–16).
   These patterns are consistent with the overall distribution of fixed
   Alu repeats in the genome. It remains to be explored whether the
   propensity of new Alu insertions in the genic region is driven by sites
   created by pre-existing Alus or is favored due to epigenetic
   differences in the vicinity of expressing genes. Insertions in genic
   regions could potentially alter the regulatory networks that are
   enriched in sites that could affect the expression of genes and
   transcripts through altered methylation, expression, editing, splicing,
   localization, etc ([145]40–43). Whether polymorphic Alu insertions and
   private insertions also harbor these regulatory sites remains an aspect
   for future investigation. New insertions have also been implicated in
   many genetic diseases ([146]44–48). In a study by Payer et al., 809
   polymorphic Alu elements have been mapped to 1159 loci implicated in
   disease risk by genome-wide association study (GWAS) (P-value < 10^−8).
   About 44 of these Alu elements were observed to be in high linkage
   disequilibrium (r^2 > 0.7) with the trait-associated SNPs ([147]49).

   The patterns of the frequency distribution of Alu insertion-deletion
   polymorphisms vary across populations and they prove as good markers
   for studying population structure and evolution ([148]10,[149]50–55).
   In our study, the frequencies of Alu insertions varied over a wide
   range from 0.04% to 99.90% suggesting different time scales of their
   insertions. The proportions of private, rare, and common insertions
   were 41.4%, 43.01% and 15.52%, respectively. Compared with the global
   datasets 60% of Alu insertions were observed to be unique to the
   IndiGen data highlighting the utility of these Alu insertions for
   understanding the genetic structure of the Indian population. From a
   set of 2232 insertions that were shared among IndiGen, HGSVC and SGDP,
   223 insertions were sufficient to cluster the different world
   populations. In a study by Rishishwar et al., on 2504 individuals from
   the 1000 Genomes project, among the 16 192 loci of genome-wide polyTEs
   (polymorphic Alu, L1 and SVA) polymorphic Alus showed the highest
   levels of resolution for human evolutionary relationships, ascribed to
   their higher diversity and numbers ([150]3).

   We could achieve 89% validation of the selected polymorphic Alu
   insertions using the PCR and Sanger Sequencing. Though many insertions
   were present in the intronic regions few of them were also present in
   the UTRs. InDel_4893 present in the 3′UTR of PTX3 gene has an
   overlapping enhancer element. The presence of polymorphic Alu in the
   enhancer region could lead to differential regulation of gene
   expression under the control of that enhancer and thus different
   outcomes. The impact of the presence of Alu elements in such regulatory
   regions would need detailed experimental studies. Alu insertions
   identified in our study are significantly enriched in genes with roles
   in the neurogenesis process. Considering that insertions in the genes
   can have an impact on gene expression and thereby its function, these
   insertions could have an impact on the neurological pathways, which
   would need further detailed validation studies ([151]28). This further
   adds to the increasing body of evidence of involvement of Alus in
   neurological diseases primarily by altering mitochondrial functions
   ([152]56). Alu insertions have been reported to contribute to
   large-scale structural variations and genome rearrangements in many
   diseases ([153]28,[154]57–59).

   Considering India’s vast population, including more samples would give
   us even better insight into the landscape of polymorphic Alus. In
   summary, this study for the first time provides a spectrum of
   genome-wide active Alu insertions in the Indian population some of
   which are shared and a majority of them being novel. This baseline
   resource would be of enormous utility in understanding population
   structure as well as identifying new disease risk loci that might be
   specific to the Indian population.

CONCLUSION

   Polymorphic Alu insertions can influence genome structure and function
   and serve as ancestry informative markers. With the recent release of
   IndiGen data from 1021 individuals, it has now become possible to study
   the genomic landscape of Alu InDel polymorphisms in the Indian
   population. This study from IndiGen adds to the repertoire of Alu InDel
   polymorphisms to the global dataset and enriches it in terms of
   diversity. This would be of enormous utility in population genomics and
   assignment of ancestry in association studies. Variability in the
   presence and expression of Alu insertion-deletion polymorphisms could
   confer population-specific differences in phenotypes and diseases.

DATA AVAILABILITY

   The Alu InDels unique to IndiGenomes have been submitted to the NCBI
   Variation Submission Portal, dbVAR with submission ID: nstd215 and also
   made available for download on the IndiGenomes database
   website [155]http://clingen.igib.res.in/indigen/.

   The codes used for InDel identification and analysis can be found on
   GitHub at [156]https://github.com/Prakrithi-P/ALU_IndiGen.

Supplementary Material

   lqac009_Supplemental_Files
   [157]Click here for additional data file.^ (2.3MB, zip)

ACKNOWLEDGEMENTS